The Economic Evaluation of Projects Economic App.pdf(Project Evaluation for the Next Decade and Notes on Some Issues in Social Project Evaluation), Glenn Jenkins (The Appraisal of

-I5 1,5 OoWIc9&

*dams Economic Development Institute*; 14,P of The World Bank

The Economic Evaluationof Projects

Papers from a Curriculum Development Workshiop

Edited by

David G. Davies

EDI LEARNING RESOURCES SERIES

EDI LFARNNG RESOURCES SERIES

The Economic Evaluationof Projects

Papers from a Curriculum Development Workshop

Edited by

David G. Davies

The World BankWashington, D. C.

/ 1996The International Bank for Reconstructionand Development / THE WORLD BANK1818 H Street, N.W.Washington, D.C. 20433, U.S.A.

All rights reservedManufactured in the United States of AmericaFirst printing July 1996

The Economic Development Institute (EDI) was established by the World Bank in 1955 to train officialsconcemed with development planning, policymaking, investment analysis, and project implementation inmember developing countries. At present the substance of the EDI's work emphasizes macroeconomic andsectoral economic policy analysis. Through a variety of courses, seminars, and workshops, most of which aregiven overseas in cooperation with local institutions, the EDI seeks to sharpen analytical skills used in policyanalysis and to broaden understanding of the experience of individual countries with economic development.Although the EDI's publications are designed to support its training activities, many are of interest to a muchbroader audience. EDI materials, including any findings, interpretations, and conclusions, are entirely those ofthe authors and should not be attributed in any manner to the World Bank, to its affiliated organizations, or tomembers of its Board of Executive Directors or the countries they represent.

Because of the informality of this series and to make the publication available with the least possible delay, themanuscript has not been edited as fully as would be the case with a more formal document, and the World Bankaccepts no responsibility for errors. Some sources cited in this book may be informal documents that are notreadily available.

The material in this publication is copyrighted. Requests for permission to reproduce portions of it should besent to the Office of the Publisher at the address shown in the copyright notice above. The World Bank encouragesdissemination of its work and will normally give permission promptly and, when the reproduction is fornoncommercial purposes, without asking a fee. Permission to copy portions for classroom use is granted throughthe Copyright Clearance Center Inc., Suite 910, Rosewood Drive, Danvers, Massachusetts 01923, U. S. A.

The backlist of publications by the World Bank is shown in the annual Index of Publications, which is availablefrom Distribution Unit, Office of the Publisher, The World Bank, 1818 H Street, N.W., Washington, D.C. 20433,U.S.A., or from Publications, Banque mondiale, 66, avenue d'lena, 75116 Paris, France.

At the time of writing, David G. Davies was principal economist in the Studies and Training Design Division ofthe World Bank's Economic Development Institute.

Library of Congress Cataloging-in-PublicatIon Data

The economic evaluation of projects: papers from a curriculumdevelopment workshop / edited by David G. Davies.

p. cm.-(EDI learning resources series, ISSN 1020-3842)Includes bibliographical references.ISBN 0-8213-3325-91, Economic development projects-Evaluation-Study and teaching

-Developing countries-Congresses. 2. Community developmentpersonnel-Training of-Developing countries-Congresses.1. Davies, David G. II. Series.HC59.72.E44E28 1996338.9'1' 072-dc2O 95-19178

CIP

Contents

Foreword v

Renewing Efforts in Project Evaluation: Workshop Summary

David G. Davies

1. Evaluation and the New Development Agenda 11

Robert Picciotto

2. On Training in Project Evaluation 17

Henry Bruton

3. Reflections on Social Project Evaluation 23

Arnold Harberger

4. Project Evaluation for the Next Decade 51

Arnold Harberger

5. Notes on Some Issues in Social Project Evaluation 71

Arnold Harberger

6. The Role of EDI in Project Training for Developing-Country Managers 83

William Ward

7. The Appraisal of Investment Projects: A Training Approach 97

Glenn P. Jenkins

iii

ForewordThe chapters in this volume were first commissioned as papers for a curriculum devel-opment workshop on Economic Analysis of Projects organized by the Economic Develop-ment Institute of the World Bank. They are now being published for a wider distribu-tion to World Bank staff and to others who are interested in updating the teaching andpractice of economic analysis.

The quality of project evaluation is a matter of ongoing concem within the WorldBank. Although the EDI and the World Bank operating departments have made pio-neering efforts to refine project evaluation methodologies and to make them more prac-tical and useful tools, recent studies of Bank project appraisals have shown that thereremains considerable room for improvement.

The Bank has addressed this challenge by providing stronger incentives for soundeconomic analysis through management and review processes, better guidelines and in-structions, and systematic staff training. Although EDI's primary focus is on the dis-semination of policy lessons, the Institute continues to support learning activities inproject analysis and management in collaboration with partner training institutes inthe developing world.

This book presents a simplified approach to project evaluation which the Bank isadopting, including how project evaluation should be modified to take into account cur-rent economic situations in most countries. The volume highlights issues that arise inapplying this methodology and how to resolve them. The two chapters that discussapproaches to teaching economic analysis will be of interest to educators and trainers.

Vinod Thomas, DirectorEconomic Development Institute

v

Renewing Efforts in Project Evaluation: WorkshopSummary

David G. Davies

The papers in this volume were contributions to EDI's Curriculum DevelopmentWorkshop on the Economic Analysis of Projects, held in Washington, D.C. on October 8-9,1992. The substance of the papers and the workshop's proceedings and conclusions havehelped EDI in its curriculum development and publications efforts. They should be ofconsiderable interest to practicing project economists, and are therefore being put into thisform for wider distribution.

The workshop was prompted by a need to respond to a demand from World Bank staffand new member countries for training in project analysis. The workshop's objectiveswere to re-examine the content of the curriculum and to review the state of the art ofteaching courses. Important considerations were (a) that EDI teaching and Bank practicebe consistent and contribute to the establishment of standard practices and terminologythat people who make investment decisions in member countries can understand easily,and (b) that EDI re-establish its leadership position in devising effective and efficientpedagogical designs for teaching the subject to professionals.

To achieve these objectives, EDI commissioned four papers by Arnold Harberger(Project Evaluation for the Next Decade and Notes on Some Issues in Social Project Evaluation),Glenn Jenkins (The Appraisal of Investment Projects: A Training Approach), and William Ward(The Role of EDI in Project Training for Developing-Country Managers). Because the twopapers by Harberger were written as follow-up and companion pieces to a previous paperby the same author, the earlier paper, Reflections on Social Project Evaluation (Harberger1985), is reprinted here. The workshop opening remarks by Robert Picciotto, directorgeneral of the Bank's Operations Evaluation Department, and the written comments onthe workshop by Henry Bruton of Williams College were valuable products of theworkshop, and are therefore included in this set of papers.

The papers and opening remarks provided a substantial basis for the workshop'sdiscussions and were further enriched by contributions by, among others, John Besant-Jones, Henry Bruton, George Psacharopoulos, Joanne Salop, and Lyn Squire. While therewas considerable debate, which is sunmnarized below, a consensus emerged with respectto the substance of practical project evaluation. These conclusions are as follows:

1

2 David G. Davies

* The use of economic surplus as the sole measure in formal project evaluationmeans that a great burden is placed on separate analyses of the social andmacroeconomic contexts of the project. Some implications are that

- Projects should be identified, evaluated, and considered only if they can beshown to contribute to a strategic plan or strategic vision. This substantiallyreduces the amount of resources required for evaluation and links the projectto its larger macroeconomic setting.

- Part of a strategic plan or vision involves establishing incentive systems thatwill govern the kinds of projects that the private sector will invest in. Theseincentive systems will include fiscal incentives as well as incentives resultingfrom infrastructure investments and regulatory changes. Decisions arerequired about how the project will take these matters into account.

- As investment projects are invariably complemented by organizationalchanges, institutional development, and other activities, assurance is neededthat these complementary activities take place and are appropriately timed.

- The recurrent costs to the government of the project should be evaluatedindependently in the context of budgetary management and planning.

* Project evaluation should be simple and understandable. This should beaccomplished by the following:

- Making growth (or preferably economic surplus) the sole measure used toevaluate projects. Projects whose objectives are, for example, to increase equityor employment, to reduce poverty, or to improve the environment are thusevaluated against their contributions or costs in terms of growth (or surplus).

- Using domestic prices at the domestic price level as the numeraire.

- Building the economic evaluation directly from the project's financial analysis,where possible.

* Changes in the relative prices of factors of production and final products requiremuch more attention in project evaluation than they are normally accorded. Priceprojections to get at changes in relative prices require at least as much, if not more,attention than the estimation of conversion factors or shadow prices.

* Computer technologies have rendered formal risk analysis accessible to all. MonteCarlo simulations should be an important part of project evaluation, especially inconnection with price projections. In many cases, point estimates of variables cannow be replaced by probability distributions and we can state the probabilities ofaccepting a bad project or of rejecting a good one.

The Papers

The main points or arguments made in each of the papers are summarized in thefollowing paragraphs.

Renewing Efforts in Project Evaluation 3

Robert Picciotto, Evaluation and the New Development Agenda

Picciotto's opening remarks essentially respond to the papers written by Harberger andJenkins. He was concerned with the narrowness of the analysis being proposed. WhilePicciotto agreed that all project economists should be skilled in economic analyticaltechniques as discussed in the Harberger papers, he believes they should also be welltrained in the use of other analytical and conceptual tools needed to deal with the majorelements in what Picciotto calls the "New Development Agenda." This agenda requiresthat individual projects be examined in the context of overall country and sector policiesand as part of the array of investment projects under consideration. In effect, this meansthat all project operations should be assessed in terms of their policy content, socialimpact, environmental consequences, effect on public finances, and institutionalsoundness.

Arnold Harberger, Reflections on Social Project Evaluation

This is a reprint of an article previously published in the World Bank's Pioneers inDevelopment series. In this article Harberger summarizes his basic approach to projectevaluation. The main feature of this approach is its strict and austere adherence to thethree basic postulates of applied welfare economics: (1) Competitive demand pricemeasures the benefit of each marginal unit to the demander; (2) competitive supply pricemeasures the opportunity cost of each marginal unit from the standpoint of the suppliers(factors of production); and (3) the difference between benefits and costs is a measure ofthe benefits or costs to society as a whole. Harberger specifically rejects the use ofdistributional weights in the analysis, which is advocated in a substantial part of theliterature, largely because weights do not represent the way citizens feel aboutredistributional efforts channeled through the public sector and their use effectivelynegates postulate (3). He advocates, instead, the measurement of "basic needsexternalities."

Arnold Harberger, Project Evaluationfor the Next Decade

This paper reviews the conditions that prevailed during the 1960s that contributed to theflowering of project evalution methodologies. These conditions included the introductionof greatly distorted prices as governments used quotas, and administered prices, licenses,and tariffs to channel or guide the future development of their countries. Hence thepreoccupation of project evaluation with correcting financial prices to their "true"economic or shadow prices. Harberger then makes the point that projections of the marketreal exchange rate over the life of a project are at least as important as estimates of thecurrent distortions in market prices.

The article provides a rather detailed, useful outline of the steps and processes ofproject evaluation. This is followed by a set of basic exercises on project timing, dealingwith successor projects, comparing projects with different lives, problems of scale, when toreplace old assets with new ones, separable components, and dealing with risk.

4 David G. Davies

Arnold Harberger, Notes on Some Issues in Social Project Evaluation

This paper is explicitly intended to review and extend the discussions in both of theprevious papers. He covers risk neutrality, portfolio risk and covariance, discount ratesand the shadow price of funds, and projections of changes in relative prices. The lattertopic is of particular interest given the author's admonition in his previous paper to payattention to prospective changes in relative prices over the lives of projects.

William Ward, The Role of EDI in Project Trainingfor Developing-Country Managers

The author, who has extensive experience working with EDI and the Bank in general,specifically addresses the question of what EDI should now be doing in project evaluation.He raises a number of issues that need to be addressed in answering this question,namely: (a) EDI tends to offer single, generic courses that are not suitable for all thevarious roles that course participants play; (b) EDI is not sufficiently concerned with theorganization and procedures of systems that might use project evaluation; (c) EDI hasneglected the technical analysis of projects in favor of financial and economic analyses; (d)EDI pays insufficient attention to important market failures as opposed to governmentfailures; (e) EDI has not integrated project analysis with policy analysis sufficiently. Forexample, an objective of project evaluation is to compensate for price distortions while animporant objective of policy analysis is to eliminate them in the first instance; (f) EDI hasyet to integrate new thinking on the role of government into project evaluation practice;(g) current Bank practice does not provide a model for training in project evaluation; and(h) many of the generic training materials EDI has developed will not satisfy specificnational needs, with the implication that EDI might consider training case writers.

Glenn P. Jenkins, The Appraisal of Investment Projects: A Training Approach

Jenkins's paper describes the course offered at Harvard University on project evaluation insome detail. Specifically, the course subscribes to the methodology of project evaluationprescribed by Harberger. The author provides a general discussion of the relationshipbetween financial and economic appraisal and the rationale for his choice of numeraire(domestic prices and domestic price levels). He then goes on to describe an intensiveeight-week course in project evaluation intended to produce professionals who canactually do the job. The course is characterized by a heavy use of staff, computers, andcomputerized case studies. The economics of project evaluation is closely integrated withpractical case work. An important feature of the course is the considerable amount of timespent on risk analysis using Monte Carlo simulations.

Henry Bruton, On Training in Project Evaluation

Bruton's paper is a commentary on the proceedings of the workshop. His comments arediscussed in the following section.


The Workshop's Discussions

Picciotto's remarks, which advocated training project economists in a broad range of areasconsistent with the considerations of the "New Development Agenda," inspired the mostimportant debate of the workshop. While Harberger and Jenkins accepted the importanceof the considerations presented by Picciotto, they argued that keeping the basics of projectanalysis as simple as possible and not burdening it with a need to incorporate objectivesother than output into the analysis was extremely important. They believed that projectanalysts should learn to do these basic techniques well and be able to communicate resultseffectively to those who make investment decisions. The techniques of project economicanalysis, they continued, were not a normal part of the tool bag of economists and couldbe acquired only with special training. In any case, the economic analysis of projects is aspecialty that engineers and accountants could learn as well as economists. Moreover,Harberger and Jenkins argued, given the need for a large number of trained projectanalysts, one could not expect all of them to have the broad knowledge and skills neededto make final decisions on projects. Both Harberger and Jenkins regard project evaluatorsas resembling accountants, who systematically go about estimating economic surpluses orlosses on projects unadulterated by other considerations.

Thus, one side of the debate argued for broad training based on the wide scope ofknowledge and analytical skills needed to make investment decisions, while the otherargued for giving top priority to the mastery of narrowly defined techniques of projectanalysis to assure that this was done extremely well. Essentially, Harberger and Jenkinsfear that a broad curriculum would be at the expense of project analysis.

Implicitly, Harberger and Jenkins assume that there has to be a division of labor.Project analysts do a narrow technical job and provide their results to others who havedifferent skills that they bring to bear on final decisionmaking. Perhaps based onexperience, Bruton, Picciotto, and Squire, by contrast, fear that narrowly trained projectanalysts would not perform well and recommend a broader curriculum. They do notenvisage that the division of labor implied in the Harberger and Jenkins model canactually be realized.

Bruton's paper, On Training in Project Evaluation, summarizes the views that heexpressed at the workshop and assesses the workshop proceedings. It also tries to find amiddle ground in the debate about the scope of the curriculum. According to Bruton, thedifferences expressed in the debate ultimately reflect differences about the sources ofeconomic growth and development, which is why he believes that agreement is sodifficult. The narrower curriculum implicitly accepts the notions prevalent in the 1960sthat growth is attributable to capital accumulation, but that attempts at rapidaccumulation lead to distortions in factor prices and the exchange rate. In the belief thatthese distortions lowered the level of investment and affected its allocation, the old schoolspecifically designed the techniques of project analysis to help project decisionmaking byestimating the social opportunity costs of productive factors (their shadow prices), andthereby eliminate the effects that distortions might otherwise have had on decisions.

The broader view, as expressed by Bruton, is that capital formation is not the basicsource of productivity growth and that a wide range of other matters should be taken intoaccount when making decisions about investments in projects. According to Bruton, theconcept of a project is actually rather ambiguous when all that is important (as discussedby Picciotto) is taken into account. Moreover, he believes that the techniques of project

6 David G. Davies

analysis are not reliable, for example, while they require projecting the values of allcomponents in a project, their record is dismal.

Besant-Jones presented data bearing on this matter. He compared Bank projections of anumber of key prices with the actual prices recorded. The differences were substantialand, what is more, seemed to be inevitable. The main lesson is that even without taking awider range of considerations into account, there will be room for considerabledisagreement on the acceptability of a project.

Squire, who at the time of the workshop was the chief economist in the Bank's Europe,Middle East, and North Africa regional office, remarked that in his region, projects wereconsidered for financing only after Bank staff had shown how they would help thecountry achieve its macroeconomic and sector policy objectives and how they would takeinto account the matters discussed by Bruton, Picciotto, and Ward. Squire pointed out,however, that the Bank's current organization did not assure that procedures wereuniform and common standards were applied with respect to project analysis. Salopconfirmed Squire's view.

Even given these problems, Bruton nevertheless regards the narrow view of projectanalysis proposed by Harberger and Jenkins as a critical development tool, and views theproject as the heart of the development process. However, he would begin any coursewith a module that puts project evaluation in the context of the country's developmentand its major problems. He would then put students through a Harberger/Jenkins type ofproject analysis. They would estimate net present values "neat," without taking equity,poverty alleviation, and other external economies and diseconomies into account. Theywould perform sensitivity analyses with respect to the critical variables. A decision toaccept or reject the project would be based on the "neat" calculation and the sensitivityanalysis.

The next step would be to consider the project in the context of the initial module,which was concerned with the country's overall economy and problems and thegovernment's policies. An answer would be sought to the questions of how the project fitsthe economy and society at the present time, what institutions are relevant to the project,and what the general macroeconomic effects of the project are. A decision should then bemade on the project's acceptability.

Workshop Consensus

The workshop participants unanimously agreed on the appropriate techniques for theeconomic evaluation of projects as spelled out in the three papers by Harberger. Themethodology is particularly valuable not only because of its clarity and meaning, butbecause it provides results in termns that decisionmakers, many of whom are noteconomists, can understand, and because students can master it relatively easily. Theparticipants also agreed that the economic conditions that project analysts should focus onin the 1990s relate to the movement of relative prices of productive factors andcommodities. Errors in price projections are far more likely than distortions to result inmistakes about projects. (The observations by Besant-Jones and Bruton should be noted inthis context.)

A number of valuable points that have direct relevance to what should be taughtemerged in the discussion of Harberger's papers. One of these was that the financialanalysis of a project should be given primacy where appropriate. Perhaps the most


important points, however, were made in Bruton's paper relating to the need to evaluateprojects in their macroeconomic contexts. Bruton's proposed curriculum seemed torepresent a satisfactory compromise between the opposing arguments on the breadth ofthe curriculum. The papers and arguments by Picciotto and Ward support Bruton'scurriculum, and are themselves upheld by Squire.

In terms of pedagogy, the participants generally agreed on the following additionalpoints with respect to teaching the narrower aspects of project evaluation:

* The center of learning should be the computer laboratory, not the lecture hall.

* Students should be taught to use a computer in the context of using a spreadsheet,preferably LOTUS 1-2-3 because of its widespread use.

* Students should learn to use Monte Carlo simulation software for risk analysisinstead of RiskMaster 2.0 as used at Harvard.

* Some form of recognition for successful completion of the course is useful, if notnecessary, for the success of the teaching program.

A Brief History of Project Evaluation in the Bank and EDI

During the late 1960s and early 1970s, under the presidency of Robert McNamara, theWorld Bank became the pre-eminent international institution for financing developmentprojects. At the same time, and to support the Bank, EDI became a virtual trade school forproject evaluation. Within its own ranks, the Bank required the use of project evaluationmethodologies through the formal technical specifications that it placed on projectappraisal operations and published in its Operations Manual. Its personnel policies gavepriority to hiring economists with a specialization in microeconomics. To establish themethodology as a basis for investment decisionmaking among the Bank's membercountries, EDI trained large numbers of government officials in project evaluation. Inaddition to improving overall investment decisionmaking by the govermnents of membercountries, the Bank expected that EDI's training efforts would result in improvedcommunications and cooperation among government project planners and Bank staffresponsible for project operations.

Project evaluation has been promoted in the Bank and EDI because it requires thesystematic collection of relevant data and information on possible investments. It helpsgovernments to avoid wasting resources at the microeconomic level and to distinguishbetween projects that are clearly beneficial from a social perspective and those whosebenefits are doubtful. It was also expected to improve communications between agovernment's central planners and sectoral or project planners on the one hand, andbetween govemments and international financial institutions on the other hand.

A number of publications reflect the contributions of Bank and EDI staff andconsultants to the development of project evaluation methodologies and to training.' Most

1. Much of the literature in the Bank and elsewhere at this time was prompted by reasoning thattoday seems simplistic: doubt was cast on the practical effectiveness of the predominantdevelopment paradigm that made a country's rate of growth dependent on its savings andinvestment rates. If countries could not save and invest sufficiently from their own resources, theymnight borrow the savings of other countries and invest them to enhance their own rates of growth,including their savings rates. This reasoning saw investment projects as the major source ofgrowth, and international lending and aid institutions, such as the World Bank, as the majorsources of other countries' savings. Naturally, planning was required to assure that aid did not

8 David G. Davies

noteworthy are the works of Tinbergen and Harberger, who both taught EDI courses.Tinbergen was a resident teacher in EDI when he wrote the Design of Development(Tinbergen 1958). For three decades, Harberger has continuously served as an EDI teacherand writer and Bank consultant. His most important written contributions include anessay on applied welfare economics (Harberger 1971) and a discussion of projectevaluation (Harberger 1972). Bank staff members Squire and van der Tak (1975) advancedthe methodology of project evaluation by showing how income distributional weightscould be taken into account. Gittinger (1982) wrote his classic and extensively usedtextbook, Economic Analysis of Agricultural Projects, and Helmers (1979) authored ProjectPlanning and Income Distribution while they were EDI staff members.

However, the Bank and EDI did not sustain the burst of intellectual and training effortson project evaluation in the late 1960s and early 1970s. From the mid-1970s onward, whileremaining primarily a project lending institution, the Bank became preoccupied withrefitting itself intellectually and operationally to deal with problems of macroeconomicand sector management. To effect these changes, both the Bank and EDI underwent majorreorganizations during the mid-1980s. The Bank retired a large number of technical projectstaff and increased the number of macroeconomists. In the process, it lost some of itssubstantial capabilities in project operations. EDI was reorganized and restaffed to handleeconomic policy and management problems, and after 1984 project evaluation courses hada low priority and were seldom offered. It was assumed that other institutions would pickup EDI's training efforts.

By the early 1990s, the Bank began to suspect that its project operations requiredrenewed attention. This led to two studies, one (World Bank 1991) on the quality of theeconomic analysis used to justify projects (the Joanne Salop Report) and the other (WorldBank 1992) on the overall quality of the Bank's project operations (the Wapenhans Report).The former was prompted by evaluations of Bank operations by its own OperationsEvaluation Department, which found that assumptions justifying projects tended to beoveroptimistic, and by criticism of project evaluation in the Bank by Little and Mirrlees(1990), who found that actual project evaluation departs from the theoretical ideal. Thisstudy found that:

e There is an enormous gap between the existing guidelines and current practice;* The economic evaluation of projects in the Bank tends to be overoptimistic and to

neglect downside risks-a tendency it shares with the Bank's macroeconomicforecasts;

* Project economic issues have been pigeonholed into a very narrow concern withthe calculation of economic rates of return, as opposed to a more fundamentalconcern with the economics of project design and strategy and their consistencywith the broad macroeconomic and sectoral strategy.

enable governments to invest in projects that wasted resources. This model, known as theHarrod/Domar model, was central to economists' thinking at the time. The much richer and moreinfluential dual economy model of Lewis (1954) and others implies a Harrod/Domar model. Thecapital-output ratio that emerges from this model figured prominently in the formulation ofdevelopment plans. A reduction in the ratio was an objective of cost-benefit analysis and projectevaluation.


The second study, the Wapenhans Report, was more far-reaching. While endorsing thefindings on project economics in the Salop Report, it found that problems in the Bank'sproject portfolio were largely embedded in the Bank's culture. Undue pressure on staff toget projects to the board led to a lack of attention to local participation in projectidentification, design, and, ultimately, in project implementation. Inadequate attentionwas paid to project supervision and to portfolio management at the country level.

In addition to these studies, which required a re-examination of project evaluation, anenormous demand emerged for project evaluation training, not only by the Bank's newermembers, such as China and the countries of Eastern Europe and the former U.S.S.R., butby more long-standing members in the Middle East and South Asia.

Training programs in project analysis are offered by Harvard's Institute forInternational Development (HIID) and at The University of Bradford in the UnitedKingdom. EDI currently supports training in project analysis and implementation incollaboration with its training partners in the developing world, especially in transitioneconomies.

EDI, however, had reasons to hesitate before embarking on renewed training efforts inproject evaluation. The economic and policy environments of developing countries andthe thinking about them had changed dramatically during the 1980s. The implications ofthese changes for the substance and practice of project evaluation were unclear.Furthermore, surveys and evaluations showed that with some noteworthy exceptions,prior efforts by EDI and others to establish project evaluation in governments throughtraining had not been as successful as anticipated. Indeed, both the Salop and Wapenhansreports revealed problems within the Bank in this respect. What were the reasons for thislack of progress? Finally, what do the new electronic technologies imply for the practiceand teaching of project analysis?

The papers in this volume attempt to come to grips with these questions and to provideguidance in making a renewed effort in project evaluation training.

References

Gittinger, J. Price. 1982. Economic Analysis of Agricultural Projects, 2nd ed. Baltimore, Maryland: TheJohns Hopkins University Press.

Hahn, F. H., and R. C. 0. Mathews. 1964. "The Theory of Economic Growth: A Survey." EconomicJournal 74, (December): 779-902.

Harberger, Amold. 1971. "Three Basic Postulates for Applied Welfare Economics: An Essay." Journalof Economic Literature 9(3).

1972. "Project Evaluation." In Collected Papers. London: MacMillan.

1985. "Reflections on Project Evaluation." In Pioneers in Development, Second Series.Washington, D.C.: World Bank.

Helmers, F. Leslie C. H. 1979. Project Planning and Income Distribution. London: Martinus Nijhoff.

Lewis, W. Arthur. 1954. "Economic Development with Unlimited Supplies of Labour." ManchesterSchool, May 22,139-91.

Little, Ian, and James A. Mirrlees. 1968 Manual of Industrial Project Analysis in Developing Countries.Vol. 2, Social Cost Benefit Analysis. Paris: OECD.

.1974. Project Appraisal and Planningfor Developing Countries. London: Heinemann EducationalBooks.

10 David G. Davies

1990. "Project Appraisal and Planning Twenty Years On," Proceedings of the World BankAnnual Conference on Development Economics, Supplement to the World Bank EconomicReview.

Squire, Lyn, and Herman van der Tak. 1975. Economic Analysis of Projects. Washington, D.C.: WorldBank.

Tinbergen, Jan. 1958. The Design of Development. Baltimore, Maryland: The Johns Hopkins UniversityPress.

World Bank. 1991. "Economic Analysis of Projects, Phase I and Phase 11 Reports." Washington, D.C.Working drafts.

. 1992. Effective Implementation: Key to Development Impact. Washington, D.C.: PortfolioManagernent Task Force.

1

Evaluation and the New Development Agenda

Robert Picciotto

Human welfare is the ultimate objective of development. It is, therefore, appropriate tobase the evaluation of development operations on welfare economics and to have ArnoldC. Harberger, a pioneer of this proud tradition, provide intellectual leadership to thisworkshop. We are also fortunate to have as participants Professors Glenn Jenkins, HenryBruton, and Anthony Bottomley.

Equally, I wish to express appreciation to Joanne Salop, Lyn Squire, and the otherBank professionals who have joined the workshop. I look forward to an insightful andspirited debate. I can think of no professional topic of greater priority or relevance thanthe one on our agenda or of a better group to deal with it.

The main theme of my remarks is that the new development agenda mandates thisgroup to bring a variety of fresh perspectives to bear on the economic evaluation ofdevelopment operations.

In architecture, form follows function-and so it is in the development business. Therole of the project economist, that is, the function of economic evaluation of investmentprojects, has changed since Little-Mirrlees and Squire and van der Tak dug thefoundations of current project evaluation methodologies. The spare elegance of theirmodels is reminiscent of Bauhaus architecture.

But we live today in a postmodern environment characterized by diversifieddevelopment objectives, multifaceted operations, and new technologies. The time hascome to reflect these requirements in the architecture of the economic evaluationcurriculum.

Accordingly, as this workshop proceeds, it will have to be responsive to today'soperational needs and to the actual dilemmas faced by our developing membercountries.

We count on the participants from academia to tap into the most recent advances ofthe social sciences. Regional staff will be called on to clarify the needs implicit in currentcountry assistance strategies. EDI's representatives will have to find ways to make thesecontributions accessible to development professionals within and outside the Bank. Insum, this workshop faces a conceptual as well as a pedagogical challenge.

11

12 Robert Picciotto

Evaluation and Portfolio Quality

From an institutional perspective, the stakes are considerable. As a financial institutionand as a development think tank, the Bank has been doing rather well. But as the leadingdevelopment lender, it has not been performing as well as it should have. Enhancedappraisal practices are essential to help improve the Bank's lending record.

Since the Operations Evaluation Department (OED) has been tracking the estimatedresults of Bank operations, the failure trend has been increasing. Whereas 15 percent ofcompleted projects approved in the mid-1970s did not perform satisfactorily, more than30 percent of those approved in the mid-1980s were rated unsatisfactory at completion.

Nor can we expect an early reversal of the trend: The rate of problematic operationsunder implementation doubled between 1981 and 1991.

Improved evaluation is equally important for our borrowers. Although it is harderand harder to generalize about developing countries, they have on average done ratherpoorly in managing their domestic and borrowed resources. While average per capitagrowth of gross domestic product (GDP) in developing countries fluctuated within a 2.6percent range from the mid-1960s to the mid-1970s, it has been stuck below 2 percent inmost years since 1986 and even turned negative in 1990 and 1991. Lessons learnedthrough evaluation should help improve the design of economic management.

Need for a New Approach

Never since its introduction as an essential tool of development financing has cost-benefitanalysis been viewed with such indifference by academics and practitioners alike.

What lies behind this dismal state of affairs? The development paradigm that ascribeda central role to planning and public investment has been discredited. Accordingly, thefocus of development economics has shifted from investment planning to policy, that is,to redefining the role of the state, adjusting economic management, and reducing publicexpenditures.

The consequent demise of national agencies concerned with planning (without acommensurate increase in the professional capacity and influence of evaluation unitswithin finance ministries or audit administrations) has had an unintended consequence:a pervasive neglect of quality assurance with respect to public investment-just at a timewhen accountable and transparent decisionmaking with respect to public expenditureshas emerged as an essential factor of structural reform and effective governance.

In parallel, overall country policy-;ather than the discrete investment project-hasbecome the central organizing principle of development assistance. The shift has beenespecially pronounced in the World Bank, which, over and above its project financingrole, has been called on to exercise intellectual leadership within the developmentcommunity; to set country performance standards for donors in conjunction with theInternational Monetary Fund; and to take a special role in the financing of adjustment.

In this context, it is not surprising that Bank economists have concentrated theircreative energy on macroeconomic and sectoral policy issues.

Evaluation and the New Development Agenda 13

Measuring Performance

The choice of appropriate evaluation methodologies is more than a matter of conceptualelegance. According to Tom Peters, "What gets measured, gets done." And W. EdwardsDeming, the guru of manufacturing quality, has issued dire warnings about theunintended consequences of accounting conventions in the corporate world.

Crossing over to development, similar warnings have come from the OED,environmental groups, and grassroots nongovernmental organizations. Their concernshave evoked a deep resonance in the development community. The rapid spread of suchconcepts as sustainability, participation, and capacity building reflects a deep uneasewith the paradigm of traditional investment analysis. Is this unease justified?

Evaluation and the Development Agenda

Development is readily defined as equitable and sustainable economic growth. Butenvironmentalists have challenged available measures of growth; equity is far from anunambiguous concept; and sustainability has at least three dimensions: financial,institutional, and environmental. Furthermore, operations are increasingly used asvehicles for broadly based policy reform and capacity building.

New objective functions animate development operations. Therefore, the tests neededfor screening programs and projects must reflect these concerns or risk irrelevance. Yetattempts to cram sophisticated adjustments into rate-of-return calculations have provenshort-lived.

Development operations are widely expected to serve poverty reduction, privatesector development, and environmental protection through improved economicgovernance and a variety of programs emphasizing human resource development, publicsector reform, efficient infrastructure creation, and effective natural resourcemanagement. We must sharpen the tools of economic evaluation to deal with theseemerging priorities.

Methodological Issues

Thus economic evaluation must, first and foremost, assess the effect of the operation oneconomic growth. But given the new development agenda, analysts must assess everyoperation in terms of: its policy reform content; social impact; environmentalconseq+rences; fInancial justification (including its impact on public finances); andinstitutional soundness

Of course, not all projects can be expected to meet all these objectives simultaneously,so tradeoffs may be involved. Are there practical ways of translating a gain in onecategory into a loss for the other? How can economic evaluation deal with multipleobjective functions?

A similar dilemma arises at the project design stage. All too often, excessivecomplexity in project design leads to implementation gridlock and to proiect failure. Sothere is a grnwing tension between the ambition of the development agenda (associatedwith a mu!tipicity -f objectives, each championed by one or more stakeholders withinand outside the developing country concerned) and implementability considerations,

14 Robert Picciotto

that is, the desirability of keeping project designs within the constraints of domesticcapacities. What are the implications for economic evaluation?

Country Focus

One implication of this conundrum is that it may not be feasible to examine the efficiency

and responsiveness of a single operation without examining the other components of thedevelopment program. This means that, beyond the individual project, economic

evaluation must concentrate on the investment portfolio. In other words, the country hasbecome a "privileged unit of account" in evaluation.

There are, of course, limits to a country-based approach:

* Growing interdependence among countries may bring a global dimension tobear on economic, social, and environmental evaluation.

* The recognition of a crucial role for private enterprise has highlighted the firm asa unit of account.

* The increasing sensitivity to social factors may bring a local, if not a household,or individual, dimension to economic evaluation, namely, in such respects asresettlement and employment effects.

Participatory Development

The advent of participatory development is an attempt to respond to this new order by

adjusting the design process for development operations. According to this increasingly

influential conception of the development process, the preoccupation with a single

measure of the net present value of a given project is less relevant than the transparent

consideration of alternative schemes by stakeholders.

In such a context of bounded rationality, effective guides to "second-best" solutions

can be more useful than cumbersome optimization models animated by a single objective

function.The expanded menu of objectives included in the development agenda explains why

it has been necessary to diversify the instruments of development assistance beyond theinvestment project. Structural and sectoral adjustment operations, sector investment andfinancial intermediation operations, and a wide variety of institutional development,technical assistance, and advisory services have been added to the standard tool kit of theBank. All of these instruments are subject to evaluation. How can evaluationmethodology deal with each of these instruments?

Sectoral Focus

Sectoral diversification must also be a consideration. The share of social sector projects isincreasing. The nature of financial sector projects is changing. Privatization, public sectormanagement, and technology development are now frequent. Freestandingenvironmental projects have emerged, and many if not most development operations aresubject to environmental impact assessments. What does cost-benefit analysis have tooffer for the evaluation of each of these projects?

Evaluation and the New Development Agenda 15

Are the rates of return assessed in one sector comparable to those in another sector?Are the methodologies applicable to an irrigation project applicable to a health project?What kind of cost-effectiveness tests are useful for social projects? Is it appropriate to usewillingness to pay as the main determinant of benefits streams in utility projects? Areexternalities suitably taken into account in traditional rate-of-return methodologies, sayfor agriculture settlement projects? How should nonrenewable resources be handled inthe evaluation process?

Factoring Out External Factors

The factoring out of exogenous variables-global and national-is an important issue forproject evaluation. For such investigations to be of practical use, however, they must begeared toward improving the engineering of development programs and projects so thatthey can withstand the intense stresses imposed on them by the macroeconomic and theglobal environment. Any serious review of evaluation methodologies should address thisconcem. In turn, what implications do risk and uncertainty have for project design andeconomic evaluation?

A Multidisciplinary Venture

In conclusion, cost-benefit analysis is solidly anchored in theory. It provides a usefuldiscipline for development project design and evaluation. But, to be relevant to the newdevelopment agenda, the framework needs enrichment by the full range of economicinquiry as well as by financial analysis and the other social sciences.

As projects become more policy laden, it will be important to assess the effectivenessof policy reform. As development instruments become more diversified, it will beimperative to trace the linkages between macro policy and microeconomics. As socialand environmental projects become more numerous, resource economics must bebrought to bear.

Improved design of institutions requires evaluaticn of operations from a public choiceperspective. Agency theory and transaction cost ecoromics may have to be tapped for theanalysis of public sector projects. Infrastructure and environmental projects may have tobe analyzed from the perspective of institutional economics.

More often than not, projects work is a struggle with the dilemmas inherent in thefinancing of (and the access to) public goods, a fertile area of recent economic inquiry.

In sum, there is no denying that development practitioners must rediscover therigorous and conceptually elegant standard cost-benefit methodology again and again. Itconstitutes the basics of the profession. It is not a question of abandoning cost-benefitanalysis. Rather, it is a question of applying cost-benefit analysis to cost-benefit analysisand being prepared to use other analytical tools and the insights of other disciplines toensure improved development effect of policy and project interventions.

The world of development has changed. More than ever, its practitioners today mustbe well versed in the basics of cost-benefit analysis. But they must understand itslimitations as well as its potential. They must be equipped with a diversified set ofanalytical and conceptual instruments to deal with a variety of decisionmaking situationsin an agile and credible fashion. To remain relevant, the EDI curriculum should constructthis tool kit. This is the challenge before this workshop.

On Training in Project Evaluation

Henry Bruton

This report summarizes the main themes of the conference and adds various commentson the themes that seem to me relevant to the training programs in the EconomicDevelopment Institute (EDI).

The conference began with a strong statement by Robert Picciotto to the effect thatproject evaluation must take place in the context of the "New Development Agenda."This requires that any project be looked at from many perspectives. Cash flows are, ofcourse, important, but so are numerous other aspects of projects. Conceptual elegance isnot the objective; rather, the aim must be to find directly usable tools that catch badprojects at the appraisal stage. The primary objective is the creation of equitable,sustainable growth, and this requires attention to policy reform, the social andenvironmental impacts of a project, and the effect on the institutions of the society. Inparticular, we must focus on poverty alleviation and employment. This is obviously avery wide view of project evaluation itself and of its purpose.

Equally strongly, Arnold Harberger has urged a much narrower view. The maintheme here is that we should seek to establish a common, widely used technique thatenables us to effectively address specific, identified things. Primary, if not exclusive,attention focuses on the effect of a project on output. We need to recognize that wecannot put prices on everything, cannot take everything into account, but we can dosome things very well. In taking these actions, we should make clear exactly what wehave done, what our assumptions are, how reasonable our dates are and so forth. Thenwe will be supplying a product that the policymaker can factor into decisionmaking withconfidence, even though all questions are not answered. It was noted that in the old days,the EDI taught a culture-it must now teach how to do things.

The difference between these two approaches rests on contrasting ideas about howdevelopment takes place, and indeed about the definition of development. The narrowerview seems to rest on the Arthur Lewis dual-economy model, which in turn had linkswith the growth model of R. F. Harrod. In the Lewis model all investment was to takeplace in the "modern" sector. The physical capital and technology that was to be createdwas the same as that in rich countries. This meant assuming that the capital output ratioin the developing countries was fixed. The Harrod model told us that the rate of growthof output is given by the saving rate over the capital output ratio. With the latter fixed,the higher the saving rate, the higher the rate of growth of output. This in turn meant that

17

18 Henry Bruton

it was crucial to allocate the new capital in such a way that the maximum growth wouldbe realized with a particular rate of saving. So project evaluation was born, anddevelopment became essentially a matter of getting a high rate of investment and thecorrect allocation of new capital. Factor prices did not matter because technologydominated, and the high rate of output growth would solve poverty and employmentproblems. These objectives would not need any independent attention.

In the 196Us economists realized that the strategy to achieve a high rate of capitalaccumulation had led to a range of policies that severely distorted factor prices and theexchange rate. Such distortion penalized both the level of investment and its allocation.Project evaluation then was directed at correcting these distortions by introducingshadow prices, and the focus was on the estimation of shadow prices. To get privateinvestment allocated correctly, the policy package was simple: eliminate the distortions-get prices right. Development then was achieved during a situation in which all priceswere right and the rate of investment high. Project evaluation, once all prices were right,would be largely concerned with projecting future prices of output, factors, intermediategoods, and foreign exchange. The narrow view of project evaluation remains strategic asa development tool, and the project is the heart of the development process.

The broad view, in contrast, emphasizes "welfare." Development is not simply moreoutput but includes a wide range of other components of welfare that must be taken intoaccount as investment takes place. When matters of distribution, health, environment,poverty, food security, values and culture, and so forth are given weight, the projecttakes on a much more ambiguous form, and evaluation becomes more a matter of theway the project fits into the general policy package than as an independent unit to beanalyzed separately. Evidently this is a much less precise notion than is the narrow view.

This wider view also recognizes that capital formation is not the basic source ofgrowth of output. Rather, one must give equal, if not more, attention to productivitygrowth. The evidence shows that productivity growth accounts for as much or more ofthe growth of output than does capital. It is, however, obvious that unless productivitygrowth continues regularly, the rate of return on capital will fall, and private investmentwill be weak. Thus capital formation and projects no longer are the heart of growth.Despite the vast literature on productivity growth, it is fair to say that economists do notunderstand its real source. In particular, we cannot say which investments lead to regularproductivity growth of the kind needed in developing countries, nor can we say whatprices are "right" to induce the search and learning that produces the productivitygrowth.

Perhaps the primary aspect of this wider approach is that it involves a great deal ofsearching and learning, of trial and error, and of recognizing the huge complexity of evena simple economy.

Much of the discussion at the conference seemed to reflect these different points ofview. I emphasize that the differences seem to rest on various views of developmentbecause it is this fact that makes resolution of the disputes so difficult. I now want toexplore these differences, as they emerged at the conference, in a bit more detail.

The Narrow View

The idea of having a rather formal, widely used technique of project evaluation thatyields an answer that is understood is an important objective. Even for those who are

Training in Project Evaluation 19

willing to accept this narrow view, there are many sources of disagreement. The mostappropriate treatment of risk is open to many questions. There are many doubts aboutthe estimation and use of shadow prices; the choice of the discount rate is bothfundamental and subject to many disputes-including whether it should be positive.There are disputes about what criterion to use in determining which projects to reject oraccept. All these questions arise even if the sole objective is increased output.

Fundamental to all investment projects is, of course, the projection of all componentsrelevant to the project. John Besant-Jones presented data that compared Bank projectionsof a number of key prices with actual prices, and the differences were generally large.Such differences seem to be inevitable. In some cases it seems possible to say that acurrent price is not sustainable and to agree on the direction that it will change. Ingeneral, however, projections are going to be wrong, in many cases drastically wrong.Calculations of expected values do not really resolve this difficulty because expectedvalues are not likely to equal actual values. Indeed, participants at the conference seemedto agree that to project more than five or so years into the future was unwise. Since manypublic investment projects are intended to last well beyond five years, it is not clear howto meet this particular problem. Thus in this narrow view different projections andvarious ideas of how to project are sure to be a source of disagreement.

Therefore, it does not seem likely, even in the context of this limited view of projectevaluation, that there will be a widespread acceptance of a particular package. At thesame time there is an important issue here, and I want to come back to it later.

The More General View of Project Evaluation

The fact that the narrow view is open to many difficulties does not mean that the moregeneral view is "right," and simple. Obviously it is not, and this larger view has as manyor more problems that must be addressed. To accept the view that matters of povertyrelief, employment, environment, deep social values, and so forth should be explicitlyconsidered creates the need to assign such objectives some sort of measure or weight, andno one knows exactly how to do this even conceptually. Many projects that the Banksupports now appear to have no present values or internal rates of return calculated. Thisseems to be due, in large part, to the fact that many projects are concerned with thoseaspects of the economy and society that simply do not lend themselves to these sorts ofcalculations. One may conclude that on these kinds of projects the Bank has determinedthat it can make decisions on a project's appropriateness without explicit calculations ofconventional measures. If such an assumption is valid for some projects, why is it notvalid for all of them? Perhaps all projects should be treated in this way. This seems to meto be incorrect, but there is an important message here, one which I will explore in amoment.

The basic question, of course, is: How does the bank decide on projects for which nonet present value (NFV) or internal rate of return (IRR) is calculated? Lyn Squire made anumber of important points on this issue. Analysts focus on the rationale of the project,of how it fits in with the general strategy of the Bank and of the country. How does itaffect the macroeconomic balance of the country? Does the idea of the project make sensein general, given the state of the country at the moment? Another important question is:Why is the private sector not already doing the project? If these questions all areanswered in a way that suggests the project looks hopeful, then attention shifts to more

20 Henry Bruton

specific matters, such as discount rates, cost recovery, effect on balance of payments, andsensitivity to exchange and conversion rates. It seemed clear that these latter calculationswere deemed less important than were the more general issues noted previously. Clearly,in this approach the traditional model of project evaluation does not really enter into thestory. As nearly as I could ascertain, the "failure rate" of projects approved in thismanner was no greater than that of those following the more traditional approach.

Such an approach puts great emphasis on general knowledge of how the economyworks, of the current state of the economy and the government, of critical bottlenecks inthe country, and so on. This kind of knowledge is very real and difficult to assignnumerical value. Someone noted that it costs one hundred or more times as much toeducate a person at the college level than at the elementary level. This piece ofinformation can be used in an illuminating way in deciding where to put the educationdollar.

Computers

Before turning to some pedagogical and curriculum issues, I want to say something aboutcomputers. Let me emphasize that I know virtually nothing about computers and amgenerally inclined to the view that the world would be better off without them. We havethem, however, and the question is how to use them in the most productive way. I haveonly two observations. Richard Lucking emphasized that the computer was a teachinginstrument, a way to help students more completely understand exactly what they weredoing during project evaluations. For students unfamiliar with computers, thinking ofthem in this way is difficult and demands careful teaching. Baher El-Hifnawi reportedthat Harvard once had one computer for each student, but then cut back to one computerper two students. The reason was that if the students had their own computers, theywould spend much too much time on them, and not enough time on other aspects of thecourse.

Both of these points are surely important, and we must find a way to exploit thecomputer without the courses becoming a course in computers rather than in projectevaluation. I suggest that it is not easy, and it is important.

Some Proposals

Given this interpretation of the conference, what makes sense for the EDI to try to do as itplaces increasing emphasis on training in project evaluation within the Bank and invarious countries? Following are some general ideas that do not exactly add up to aformal proposal but may contain ingredients of a more complete outline:

1. There should be an opening session of, say, one morning, or one day that puts theproject evaluation exercise in the context of the entire development objective. This shouldbe done in a careful, systematic way and should recognize the extent of disagreement inthe profession, the range of objectives, the state of knowledge and ignorance, and somethoughts on how our views have changed since the 1950s. I put great weight on doingthis as objectively as possible.

This discussion would identify the main problems of the country at the present time.These might be employment, poverty, inequality, and the environment, but they might

Training in Project Evaluation 21

be something more nebulous, such as entrepreneurial shortages, lack of response of theprivate sector to price or profit incentives, and corruption. These should be specificallyidentified. In this context, it would be possible to talk about the origin of projects, animportant topic not frequently discussed.

2. Then the group could work through one form of the narrow version of projectevaluation. The obvious one to use is that of Harberger, the most convincing (in myview) one available. The group could use only output as the objective, that is, nothing onpoverty, environment, or other problems. It should be fairly easy to introduce some kindof sensitivity analysis into this: What happens if the cost of foreign exchange is 5- less or5- more than it is? What happens if the discount rate is moved around a bit?

The class should spell out assumptions made at every step and specifically state therationale for making them. Much of the class discussion would center on why theseassumptions, not others, were made. A few hypothetical examples should be used at theoutset, but students should be put in touch with "real" projects early. My view is thatparticipants in such a course should do their own projecting, choose discount rates, andso forth rather than assume that some other office does that. To depend on someone elsefor the projections, lets the project evaluator off the hook too easily-given the problemslisted above. Once an NPV is obtained, the assumptions specified, and the sensitivitywork done, the class will make a decision to accept or reject it.

When projects do not lend themselves to this sort of treatment, the class shouldrecognize this explicitly. It may, however, be possible to gather data that are relevant tothe issue of such a project, for example, education costs, as mentioned above. It would beunfortunate, in my view, to teach that it is possible to calculate the NPV of a projectaimed at reducing, for example, infant mortality even though everyone agrees it isadvantageous to reduce infant mortality.

3. The class would then study this result and its justification in the context of thematerial covered in the first session. The basic question would be one noted earlier: Howwell does the project fit the economy and society at the present time?

(Let me illustrate this latter point with an example from one of Al Harberger'scomments. He mentioned that a group of economists from Cambridge had asked him tosupport their proposal of a $30 billion aid package for Russia. No information wasavailable on what the money was to be used for or how the total was determined. Herefused to lend his support to the proposal. He was surely right to refuse. Then Al said ifthe group had listed some projects, things would be different. For example, we all knowthat telephones are a necessity in a modem society, and a major aid program to providetelephones in Russia could well be accepted. This may be, but I would doubt it. Given thepresent disarray of the Russian economy, I doubt that telephones would break manybottlenecks or would fit in with the economy. I would believe this, even if the NPV of atelephone project was positive.)

To do this part of the exercise well requires a lot of intuition. There are people indeveloping countries who have a lot of intuition. We need to find a way to exploit thissource of information. This is difficult. People with good intuition are hard to find, hardto get information from, but they are there. Part of the exercise at this point might includea discussion of who would have information about a project's effect and how to get themto tell what they know. Certainly this source of information should be recognized.

After this is done, the class should review its decision to reject or accept. If it changes,a complete statement of why should be spelled out.

22 Henry Bruton

It may be noted explicitly that this approach does not involve putting weights ornumerical values on the environment, poverty relief, and so forth and then includingthem in the calculation of NPV. Rather, the NPV is calculated "neat," and then thedecision that emerges is studied in a larger, more inclusive context. It is studied no lessrigorously or formally, but in a different way. The class would use data but not in theexplicit way done in the conventional project evaluation form.

One might also consider the difficulties arising from the fact that most projects areexpected to last longer than it makes sense to project their prices, costs, and so on. Howcan that be taken into account? I think the main policy implication of this situation is torecognize the advantages of considerable flexibility. The least-cost method may not bethe least cost if it results in an inflexible, unadaptable piece of physical or human capital.Indeed, in an era when technology is changing rapidly, flexibility and adaptability are ahigh-priority objective.

4. Participants frequently mentioned "institutions" at the conference. This is a difficultnotion to define. In a course with participants from different countries, it is often effectiveto ask them to identify specific institutions in their country that seem relevant toappropriateness of a project. It is important to try to make the student be very specific.Only then does an emphasis on institutions have a great deal of empirical content andpolicy relevance.

5. The last part could be a study of the general macroeconomic effects of the project.6. The final result should be a specific decision to reject or accept. It does not seem

useful for the project evaluator to say to the minister or policymaker, "Here are theeconomic aspects of the project, now you decide." Project evaluators know more thanministers are ever likely to know-and may be more objective as well. They should bewilling to accept the responsibility. It will help make them more careful.

Pedagogy

Everyone seemed to agree that the classes should beformally conducted. Participantsshould have graded tests, graded homework, and get a final grade. It should be madeclear at the outset that it is possible to fail. The basis for this emphasis is that participantsmust be motivated-and motivated strongly-to go all out to learn the material.

I agree completely with all this, but I do want to turn it around a bit. I want toconvince the participant that the material is important, vitally important, it really matters.Make this clear on the first day. Then the discussion of how to do the job makes moresense and (I would hope) provides the real motivation. The testing then would bejustified by the need to ensure that the participants understood the material, really andtruly understood it. The tests would not be simply to see who failed and who passed, butto be sure that everyone passed and understood . . . and they understood because theyrecognized that such understanding was of great relevance to their country.

3

Reflections on Social Project Evaluation

Arnold Harberger

I approach with considerable humility the task of organizing and expressing thesereflections because I do not consider my work in the field of project evaluation to becharacterized by great originality. What I have strived for over the years is betterexpressed by the word "professionalism."

This statement applies at two levels. First, there is the level of that largely unsung hostof people serving in budget bureaus, planning authorities, and all types of ministries andagencies (among them the World Bank itself) all over the globe, who strive selflessly tosee to it, insofar as they can, that projects not meeting adequate standards are rejected,while those in the social interest are accepted. These honorable people (for I exclude themany who demean project evaluation by using it as a device for "justifying" whateverprojects their clients or superiors want) must literally number in the thousands. They, inmy view, form a nascent profession. To me, the task of helping them develop thisprofession lies more in distilling the knowledge, wisdom, and common sense that arealready part of our heritage than in extending the frontiers of knowledge in any deepsense. Moreover, I have always thought that in doing so one should respect the goal ofsimplicity as much as possible-seeking principles and rules that can be readilycommunicated to and understood by this heroic (and hopefully growing) band ofpractitioners, most of whom belong to professions other than economics.

This brings me to the second level at which the word "professionalism" is relevant.The economics profession has been with us for a long time. The insights and wisdom ithas accumulated over more than two centuries are probably the most important sourcefrom which to derive the principles and rules to guide the new profession of socialproject evaluation. This is the star that I have tried to follow as I have dealt over the yearswith project evaluation: to distill from the huge corpus of economic science the conceptsthat were particularly relevant for the new profession, and to function as a professionalmyself (rather than as a scientist) in studying different areas of application.

This same spirit motivates the present paper. In the course of it I will surely have todisagree with other professional colleagues who have dedicated their time and efforts towork in the same vineyard. But even here the spirit of the disagreement is betterdescribed by "Who has found the simplest and most convenient distillation of ourprofession's accumulated tradition?" rather than by "Who has invented what?" or even,on a given point, by "Who is right and who is wrong?"

23

24 Arnold Harberger

Three Basic Postulates of Applied Welfare Economics

The grand tradition of applied welfare economics, going back at least to 1844 and thedays of J. A. Dupuit (1952), can be interpreted as being based on three simple postulates:

1. Competitive demand price measures the benefit of each marginal unit to thedemander.

2. Competitive supply price (or marginal cost) measures the opportunity cost ofeach marginal unit from the standpoint of the suppliers (factors of production).

3. In attempting to measure the benefits and costs to a society (or group) as awhole, one must take the difference between benefits ( + ) and costs (-).

On the basis of these postulates, we can derive the classical principles of economicpolicy analysis. Among these are (1) the measure (a triangle between demand and supplycurves) of the so-called welfare or efficiency cost of a tax; (2) the traditionaldemonstration of why the exercise of monopoly power is not in the social interest; (3) thecase for the so-called optimum tariff or export tax, which shows how a nation (althoughnot the world as a whole) can benefit from exploiting its monopoly power, if any, overproducts it imports and its monopoly power, if any, over products it exports; (4) thegeneralized expression for the welfare or efficiency cost of taxation in a general-equilibrium setting with many commodities, developed independently by many authorsbut perhaps most elegantly by Hotelling (1983); (5) various rules for optimal taxation, ofwhich the so-called Ramsey rule (1927) is among the most famous; and (6) the so-calledLerner theorem (1936) showing the equivalence of a uniform import tax with a uniformexport tax under conditions of balanced trade.

These examples are impressive because of the power of the ideas involved and theirsignificant place in the accepted corpus of economic policy analysis. But truly they areonly the tip of the iceberg. The tradition of what is called consumer surplus analysis hashad many distinguished representatives, among whom Alfred Marshall (1890) and A. C.Pigou (1932) stand out as giants.

The three basic postulates have sometimes been criticized as not yielding in everycircumstance "true" measures of how the utility of individuals changes when somepolicy is introduced or some other disturbance occurs. But as a basis for the actualanalysis of real-world policies and projects, the postulates have not been seriouslychallenged, let alone surpassed.

Some authors have worried about cases of multiple equilibria, others about examplesin which the postulates need not always yield the same measure if a sequence of policiesis imposed in a different order (the so-called integrability condition). But I do not knowof even a single case where the presence of multiple equilibria has been identified as afactor in an important real-world policy problem. And as for the integrability condition,my favorite analogy is that we all know that the distance between two cities will varywith the route we take. But most road maps, geography books, and airlines present onlya single number for distance. The logical answer in the case of intercity distance issometimes "as the crow flies," sometimes "the shortest available route." Thecorresponding answer in applied welfare economics (when the sequencing of policysteps is an issue) is to choose the most plausible or likely sequence or, if there is none, to

Reflections on Social Project Evaluation 25

assume that all the policies are imposed together (in technical jargon, to assume "a radialexpansion of the vector of distortions" from the relevant starting point to the end result).

From my own standpoint, I have always thought of the three basic postulates asexactly that-simple postulates on which economists have constructed a system ofmeasurement (traditional applied welfare economics). All economists know that nationalincome and gross national product are inaccurate measures of national welfare. Yet moststudies rely on them. In spite of their defects, they have performed reasonably well inmost contexts. What can be said of the three basic postulates is that they are considerablymore subtle and more refined than the rules on which national income accounting isbased-they would not, for example, make the mistake of implying that welfare fallswhen mothers voluntarily leave the labor force to take care of their homes and families.In every case that I know of where the three postulates lead to results different fromthose derived by the rules of national income accounting, the postulates (as in the aboveexample) win hands down. For the real-world problems they have been used to solve,they have proven more adequate than anything else we have. They are thus the naturalsources from which to draw the principles to guide the budding profession of socialproject evaluation.

On Distortions and Externalities

Using the three basic postulates makes it easy to understand the sense in which the grandtradition of economics has always looked on an undistorted and fully competitiveeconomy as an optimum. If demand price as seen by demanders in each market is equalto supply price as seen by suppliers, and competition prevails, marginal social benefits asmeasured by postulate 1 will always be equal to marginal social costs as measured bypostulate 2. This happy state no longer prevails when distortions or externalities arepresent.

Distortions can take on many forms, of which the simplest to analyze are taxes andsubsidies. When these are present, marginal social benefit as measured by the price paidby demanders differs from marginal social cost. If we represent the excess of demandprice over supply price in an activity by Di and the change in the level of that activity byAXi the introduction of a new policy or project will produce net benefits or net costs,through its effects on other markets, according to whether the expression Xi Di AXi ispositive or negative.

I have always thought of Di as standing for "distortion," with the term construedbroadly. That is to say, the above expression is valid generally-not only for taxes andsubsidies but also for the distortions present in more complicated public policies and forthose stemming from monopoly and monopsony situations. Externalities that vary withthe level of an activity, such as traffic congestion with the volume of traffic or smokepollution with the volume of a factory's output, can also be treated as distortions (Di)within the same simple formula.

All this is important because, as will be seen below, these distortions are preciselywhy (within the framework of the three postulates) we have to build a system of socialproject evaluation that is different from the simpler economics of a distortion-free world.In particular, the pervasive presence of important distortions is the main reason it wasnecessary-or at least exceedingly useful-to construct these concepts.

26 Arnold Harberger

On Social Opportunity Costs in a Market Setting

As pointed out previously the postulates would assert that in the absence of anydistortions in the economy, the social opportunity costs of marketed goods and serviceswould equal their market prices. When new demand for a good or service is generated(say, by a new project), there are only two sources from which to satisfy that demand:increased total supply and the crowding out (displacement) of other demanders.Postulate 2 tells us that the increased supply should be evaluated at the supply price,while postulate 1 tells us that the displaced demand should be evaluated at the demandprice. In the presence of a distortion, say, a tax, these two prices are different and thesocial opportunity cost becomes (at least in simple cases) a weighted average of them.Another way of expressing this is that demand price Pd is equal to supply price Ps plusthe distortion Di. If the social opportunity cost of one unit of the good or service is aweighted average, fi Pd + f2 ps, this can be expressed as (fi+f2) Ps -4f Di. Butf1 issimply the measure of the amount of displaced demand, so the above formula boilsdown to saying that the social opportunity cost of a unit can be expressed in the formPs +Di AXi.

The above is the most rudimentary example of a general rule. Other cases are morecomplicated but come down to the same thing. A simple capital market example containstwo taxes, a capital income tax, tc, and a personal income tax, tp. There are thus threerates of return: p, the gross of tax return to investment; i (= p - tc), the market rate ofinterest; and r ( = i - tp), the after-tax return received by savers. In this market, fundsraised by a project come in part from displaced investment, which by postulate 1 has anopportunity cost of p, and in part from newly stimulated saving, which by postulate 2has an opportunity cost of r. The social opportunity cost of capital can then be expressedas fl p + f2r. But this is also equal to f1 (i + tc) + f2(i - tp), so given thatfi +-f2 = 1, it is equal to i + ftc -t2tp-that is, it is a market rate of interest adjusted fordistortions by the Li DiAXi principle.

No matter how many sectors we add, the fact remains that it is substantiallyequivalent in such cases to look at social opportunity cost as a measure that is a weightedaverage of the demand prices pd of those sectors whose demand was displaced by theentry of new demand, and of supply prices ps of those sectors whose supply wasstimulated, and as a measure that is a market price (like ps in the first case or i in thesecond) adjusted by a factor based on the application of Li Di Axi. Here AXi representsprecisely the same displacements of demand and stimuli to supply incorporated in theweights of the weighted average. Perhaps the most elegant case is the representation ofthe social opportunity cost of foreign exchange in terms of a weighted average of a hostof individual terms, which reflect the different tariff treatment of many categories ofimports and the various taxes and subsidies applying to different categories of exports.For a dollar's worth (the dollar being the main foreign exchange unit for most smallcountries) of imports of good j the demand price (postulate 1) is Em + T1, and for adollar's worth of exports of good k the supply price would be Em + Sk. Here Em is themarket exchange rate for the dollar, and Sk the subsidy per dollar of exports of k. If thereare many categories, we have weights of fj and fk reflecting the fractions of a dollar newlydemanded (say, by a project) that are brought about through displacing imports of j or bystimulating an increment in exports of k. The end result of this exercise is the followingexpression for the social opportunity cost of foreign exchange:


Ijfj(Em + Tj) + Zkfk(Em + Sk.

This is a weighted average of the demand prices (postulate 1) of many classes ofimports and of the supply prices (postulate 2) of many classes of exports. But sinceLjfj + Zkfk = 1, the expression above can equally well be written as

Em + XjfjTj + LkfkSk

This takes the form of a market rate of exchange adjusted by the Li Di AXi principle.In all the above cases the weighted average of demand prices for displaced demand

and of supply prices for newly stimulated supply (based on a quite intuitive applicationof postulates 1 and 2) is equivalent to the representation of social opportunity cost as amarket price, duly adjusted by the same weighted average of the relevant distortions(based on the Li Di AXi principle, a less direct but more subtle application of postulates 1and 2).

This leads us to see that the standard weighted-average measures of socialopportunity cost can be regarded as attempts to follow an approach that (though subtleand correct) asks us to look at the effects of a given action on all the distorted activities ofthe economy, and to assemble the adjustments for distortions into convenient"packages." For the weighted-average measure, the elements in the package are thesupplies and demands in the various component parts of the given broad market. Thus,for the social opportunity cost of foreign exchange, the weighted-average measureincludes the supplies of all the various export categories and the demands for all thevarious import categories, while for the social opportunity cost of capital it wouldinclude the demands for all the various investment categories and the supplies of savingsfrom different sources (presumably with different marginal tax rates, hence differentsupply prices of their savings). Of course, in all these cases, we can lump severalcategories together if they have the same, or closely similar, distortions.

But there may be times when the relevant package includes elements above andbeyond those in the weighted average. In deciding on the relevant package, one mustrecall that in determining social opportunity cost we are attempting to trace whathappens when new demand enters a market. Weighted-average methods trace out thoseconsequences just on the constituent parts of that market-displacing imports andstimulating exports, for example. These displacing and stimulating effects work throughthe real exchange rate, and one must recognize that there may be other consequences. Ihave suggested, for example, that if a rise in the real exchange rate ends up displacingsome imports of petroleum, one side effect would be reduced receipts from gasolinetaxes. I have also suggested that if there was a tax on bricks, and if the introduction ofnew demand into the capital market displaces construction, there would similarly be aside effect in the form of reduced receipts of that tax. When these effects are relativelyminor, the principle of simplicity suggests ignoring them. But when they are significant,they should be taken into account. Most important, perhaps, from the standpoint of thepresent paper, is the clear conclusion that weighted-average measures of socialopportunity cost are not in themselves the correct solution. In a market situation they willalways be a component of the correct solution, but they may at times need to besupplemented in important respects. All of this, of course, follows directly from the threebasic postulates that have guided applied welfare economics from its beginnings.

28 Arnold Harberger

Some Additional Observations

Let me set the stage by pointing out an important characteristic of a well-functioningcapital market or foreign exchange market. Particularly in the latter case, one can almostguarantee that the reaction will be the same: The real exchange rate will go up by thesame amount, and the same set of displacements of imports and stimuli to exports willoccur- regardless of who is the buyer of a given amount of foreign exchange. In the caseof foreign exchange in well-organized markets, the names of the buyers and the purposesfor which they buy are not even known by the market, most purchases being made byfinancial institutions or other intermediaries. The market simply "feels" the pressure ofan added demand, and a set of market-determined reactions ensues.

Some people dispute the above proposition by saying that it does not take intoaccount how the foreign exchange will be used, and that when this use is considered, thesocial opportunity cost of foreign exchange will vary from case to case according to theattributes of the use. I believe that this line of reasoning misses the whole point of theconcept of the social opportunity cost of, say, foreign exchange. The main reason forassembling certain XjDiAXi in packages is that these packages turn up with greatfrequency. When this is so, one can calculate the package solution just once and save lotsof time and effort.

This, I believe, is what we accomplish when calculating the social opportunity cost offoreign exchange. To my mind, the social opportunity cost of foreign exchange (in amarket setting) has no relation to its use. To put it graphically, if an enterprise orindividual in country A enters the market, buys dollars, and then suffers a fire in whichthe dollars burn up, what is the loss to the country? This loss is measured by the forces ofimport displacement and export stimulus and is the same regardless of who was thebuyer or what intentions he might have had about the use of the now-lost dollars. Here isa perfect example of how social opportunity cost can be separated from use. Of course,one cannot neglect the use of foreign exchange in social project evaluations. But here, bythe very nature of the case, a different pattern of effects (XDiAXi) in distorted marketswill take place in almost every operation. Foreign exchange spent on highly taxedimports (Di positive and large) will produce an important benefit, which should beattributed as a benefit of the project. If spent on subsidized imports (Di negative), it willproduce a cost. Application of the three basic postulates requires that these benefits andcosts be taken into account in the analysis of each project. But this does not annul theusefulness of the package represented by the social opportunity cost of foreign exchange.Quite the contrary, when a project buys foreign exchange, the package is a cost, and theanalyst need examine only the DiAXi that are involved in spending it. Similarly, whenforeign exchange that has been earned is sold in the market, the package is a benefit andthe analyst need examine only the DiAXi that occurred as a consequence of the particularactivities by which it was earned.

The social opportunity cost of capital basically shares the same attribute. Again, whenmoney is drawn from a reasonably well-functioning capital market, people rarely knowfor what purpose. But the withdrawal of funds tightens the market and produces apackage of responses that is likely to be very similar, regardless of who was doing thewithdrawing. There is a difference between the foreign exchange market and the capitalmarket, however. In the former, the price paid (say, for a dollar of foreign exchange) isessentially the same for every buyer, and so are the market reverberations that produce


the difference between the social opportunity cost of foreign exchange and its essentiallyuniform (among buyers) market price. In contrast, the rates of interest paid by differentclasses of borrowers vary substantially, but the market reverberations that follow fromthe withdrawal of funds from the capital market are likely to be virtually the same,regardless of who withdrew them. To put it simply, the lender learns who the borrower isand charges him a supply price; but the market, which then reacts to the greater scarcityof funds, does not know who the borrower is. Its reaction is governed only by the factthat funds are tighter.

Thus the reverberations (TiDiAXi) may cause the social opportunity cost of capital tobe 4 percentage points above the market price, but the market price might be 8 percentfor the government, 10 percent for a good industrial borrower, and 12 percent for anormal commercial borrower. In this case, the social opportunity cost of funds would be12 percent for the government, 14 percent for the industrialist, and 16 percent for thecommercial borrower.1 Most actual social project evaluations deal with public sectorprojects, for which the government rate is appropriate. In some developing countries,however, there have been serious analyses of the social costs and benefits of projectsplanned by the private sector, and in such cases the appropriate procedure is to adjust aprivate (real) cost of borrowing (10 percent in the industrial case above) by a premiumthat takes into account the reverberations (MiDiAXi) from the borrowing.

Furthermore, the time should come when students of the development process takeseriously the idea of making extensive postevaluations of both public and privateinvestments. When this happens, it will also be appropriate, in principle, to consider thesocial cost of capital to be different for broad classes of private borrowers.

As in the case of foreign exchange, there is going to be an entirely different set ofreverberations (1jDiAXi) stemming from the way the money is spent. These have nothingto do with the market for capital funds and should appropriately be taken into account inanalyzing the outlays of each project.

The Social Opportunity Cost of Labor

The application of the three basic postulates to the calculation of the social opportunitycost of labor is vastly more complicated than the applications in the preceding section.There are typically huge variations in the wages of labor according to occupation andskill. Even within these categories in many countries there are substantial regionalvariations in wages. One must also deal with the fact that individuals can and do moveamong occupations, regions, and jobs and that they have demonstrably different supplyprices in at least some of these cases.

The Important Role of Supply Price

To continue with the three basic postulates, the starting point for measuring the socialopportunity cost of labor must surely be postulate 2. This has powerful implications. If aworker is willing to work at different jobs but demands more pay for some of them than

1. These rates should in principle be defined net of the actuarial component of default risk,which is not part of the social cost of borrowed funds. But in fact rates do differ by these amounts,and more in many cases, even after correction for the actuarial component.

30 Arnold Harberger

for others, he does not have a single opportunity cost of labor; instead, he has as many(private) opportunity costs as he has asking prices. But for any one job, only oneopportunity cost is relevant-his supply price for that job.

Obviously, it is folly to think of conducting real-world social project evaluations byfinding out the supply prices of all the workers hired. The key lies in the presumptionthat, in the absence of clear evidence to the contrary, the project is paying market wagesfor each job, in each skill, and in each occupation. This in turn leads to the presumptionthat for most workers the wage represents or approximates their supply price,determined mainly by what they could get for an equivalent job in the same market.Some workers may be willing to take their jobs for less than the market wage and maythus be earning economic rents. But if all (or in practical cases nearly all) the workers on aproject are earning economic rents, that project simply cannot be paying market wages-it must be paying above-market wages, a case I will take up later.

This clearly implies that analysts must treat the social opportunity cost of labor as anessentially microeconomic phenomenon, at least in social project evaluation, althoughpossibly not in cases of major macroeconomic adjustments. The prevailing market wage,although different among types and categories of labor, must reflect the market clearingsupply price of labor under postulate 2.

Yet this supply price is far from being the social opportunity cost of labor. To get tosocial opportunity cost, we must follow a path directly analogous to the ones pursued inthe case of foreign exchange and capital. We begin with a somewhat semantic point.When a worker states his asking price (what he expects to be paid), he typicallyrecognizes that he will have to pay income tax out of the proceeds. In this respect themarket wage, wm, is similar to the market interest rate, i, presented above in analyzingthe social opportunity cost of capital. The true supply price of labor, ws, must be definednet of taxes (at least in applied welfare economics) because it is only out of net incomethat workers presumably gain their welfare and satisfaction. If we take the market wageWm as the base, we must recognize that there is a distortion tp (personal taxes), lyingbetween Wm and the true supply price of labor, w5, just as a similar distortion liesbetween i and r in the case of the social opportunity cost of capital. 2

The easiest way to think of the social opportunity cost of labor in a given area,occupation, industry, or other category is to start with the market wage and make a seriesof adjustments. All the labor of each type earns its relevant market wage. The firstadjustment is to recognize that taxes, tp, are being paid on those wages; this represents apositive external effect of the form DiAXi. But now we must turn to the ultimate sourcesfrom which that labor was drawn. Some may have been drawn from other regions, somefrom other industries within the same region, some from competing firms within theindustry. In addition, there will be some who newly joined the labor force, and at least in

Z I hope readers will forgive my using the term "supply price" sometimes in a gross-of-taxsense and at other times in a net sense. To use "asking price" as a technical term is, I believe,stretching it too far, but when it does not seem stilted or possibly confusing, I have used it. In othercontexts I have stuck with "supply price." I trust that in such cases it will be evident from thecontext whether I am referring to the market wage, wm, or the net-of-tax wage, ws.


times of less than full employment there may be some net reduction in the number ofunemployed.

For all but one of the above sources there will be some sort of external effect of theform DiAXi. All three of the first groups will have ceased to pay income taxes in theirformer place. These lost taxes represent a negative external effect of the project. Theymay just cancel the positive effect of taxes paid by the workers on the project, but thisshould not be presumed. Wages may have been higher or lower in the source than in thedestination; in countries organized on a federal basis, taxes may also differ from region toregion. When it comes to displacing labor from other industries, we must consider taxesthat may be levied on the products of those industries. These taxes, too, are part of themarginal product of the factors of production in those industries, for the marginalproduct of the factors must be valued at the prices paid by the ultimate demanders ofthose products (postulate 1), but their loss may be partially offset if other factors end upsubstituting for the labor that was in net terms drawn away. Labor drawn from the ranksof the unemployed is likely to produce an external benefit. In cases where unemploymentcompensation is being paid, any net reduction in the number who would probably beunemployed in the absence of the project means a corresponding saving of public funds,presumably to the benefit of the citizenry at large. This saving is an obvious externalbenefit. It is to be expected (and indeed has been demonstrated many times) that thepresence of unemployment compensation causes the asking price of the unemployed tobe higher. Where there is no unemployment compensation, the absorption of theunemployed creates a benefit in the form of economic rent (consumer surplus) thatreflects the excess of the wage received over the true supply price of the unemployed(which is unlikely to be zero, as was so often assumed in the early developmentliterature).

The vicissitudes of measuring the social opportunity cost of labor are vast, as indeedare those associated with foreign exchange and capital. In practice, the actualmeasurement is always immensely cruder than the underlying concept. I have tried toshow that even though the task is more complex and differentiated for labor than for theother two, the three basic postulates lead to essentially the same general solution to theproblem of measuring social opportunity cost. Whether hiring labor, borrowing capital,or buying foreign exchange, we must first determine (estimate, guess, or assume) thepattern of the ultimate sources of supply. That pattern will inevitably entail somesourcing from new supply and other sourcing from displacement of other demand. Insimple cases one can merely assign the net supply price as the opportunity cost of newsupply and the gross demand price as that of displaced demand. In more complicatedcases, which the labor market exemplifies, the more subtle approach of working with amarket price corrected by distortions is indicated. In the case of labor we start with themarket wage and correct it for all relevant externalities: taxes forgone in the varioussources, unemployment compensation saved, and consumer surplus gained by newlyhired workers. We also need to remember to introduce a credit for income taxes paid bylabor at the destination. In this respect the procedure differs slightly from that of foreignexchange and capital, where all distortions were to be treated in the explicit analysis ofthe benefits of each project. The reason is that within a labor market category (region,occupation, or some other), which we define by the fact that roughly a single marketwage applies, the relevant tax rate, tp, will not vary from project to project, and simplicitycalls for incorporating the adjustment directly into the cost of labor.

32 Arnold Harberger

The Social Opportunity Cost of Labor with a Protected Sector

A final complication to the labor story is the frequent existence of what I call "protectedsectors." Others have spoken of duality in the labor market, of traditional versus modemsectors, and so on. I like the term "protected sector" because it directly connotes a wageabove what would be a free-market equilibrium level. It also connotes that some element(such as minimum wages, union bargains, or "political insurance" by multinationalcompanies) must be at work to create the differential.3 In addition, the protected-sectorconcept, more readily than its alternatives, opens the door to the idea of not just two butquite a number of protected sectors, ordered hierarchically according to the wages theypay for equivalent work.

Another phenomenon in conjunction with protected sectors is that of quasi-voluntaryunemployment. Consider that there is a given protected-sector wage (wp) and a free-market wage (wf). The protected sector cannot hire all who want to work there, for to doso would mean bidding the market wage up all the way to wp, in which case thedichotomy between the two sectors would cease to exist. With a protected sector ofmoderate size (compared with the size of the relevant labor market), the normal labormarket structure is one in which a significant number of people having supply pricesbetween Wf and wp were not lucky enough to be employed at wp but have no interest inworking at Wf. These are the quasi-voluntary unemployed. In such a situation anexpansion in the number of jobs in the protected sector will draw workers in part fromthe free-market sector and in part from the quasi-voluntary unemployed. The averagesupply price, and hence private opportunity cost (postulate 2), of those who fill theprotected-sector jobs will accordingly be above the free-market wage. Quasi-voluntaryunemployment is thus an institutionally induced phenomenon, which curiously entailsan opportunity cost of the unemployed that is above rather than below the free-marketwage.

If the setting of the protected sector is the urban labor market, and if there is readymigration from rural areas, it is likely that the presence of the protected sector will inducerural-urban migration. This has the effect of driving down the free-market wage andswelling the ranks of the quasi-voluntary unemployed. It is really a case of migration-induced unemployment, as noted early by Lewis (1984) and analyzed in some depth byHarris and Todaro (1970), by myself (Harberger 1971), and by others.

Migration-fed unemployment is the result of the phenomenon of rent seeking, whichin turn is easily rationalized in terms of postulates 1 to 3. Wages offered at the destinationare above those required to induce migrants to move. The normal workings of an openlabor market would entail a fall in wages as increased supplies of workers makethemselves available. That is to say, the wage rate itself would serve the function ofequilibrating the labor market and in doing so would stem the flow of migration. Whenwages are maintained at protected levels and are prevented from adjusting, labor marketequilibrium is nonetheless brought about in some other way. In the case of migration-fedunemployment, the unemployment rate itself brings it about. One starts with a supply

3. The protected-nonprotected distinction does not refer, for example, to high-skilled asopposed to low-skilled jobs or to experienced as opposed to inexperienced workers. It refers toworkers of identical characteristics and abilities receiving different rewards in essentially identicaljobs in two sectors of the same labor market.


price of migration (a wage at destination that would just barely induce the migrant tomove). Initially, the wage at destination is above this supply price, and the migrationoccurs. But as unemployment at the destination mounts, the supply price of migrationincreases. That is, in the presence of greater unemployment, the same destination wagebecomes less attractive, and the incentive to migrate declines. With enoughunemployment the tide of migration is stemmed; unemployment at the destination,however, remains high as a continuing equilibrium phenomenon as long as the wagepaid at the destination remains above the supply price of migration that would prevail inthe absence of unemployment.

I like this particular way of viewing the adjustment process via migration-fedunemployment because it is framed directly in terms of supply prices and demandprices. The demand price is the fixed wage at the destination; the supply price is thewage that would just barely induce (or compensate) migration. As unemploymentmounts, this supply price rises, and the adjustment becomes complete when this supplyprice equals the fixed destination wage.

This demonstrates very simply the usual result: that under conditions of migration-fed unemployment the social opportunity cost of labor will end up being equal to thefixed destination wage. Raising the latter would only increase unemployment until thesupply price of migration again equaled the wage. Not only do we reach this resultsimply, but once again we see the power of the three postulates. The answer is aprofound expression of postulate 2. If demand price exceeds supply price and is notallowed to adjust, then something else (here unemployment) will develop so as to bringsupply price up to demand price.

Social Opportunity Cost in Nonmarket Situations

Except in the preceding section, I have dealt with a world of market prices-prices thatmay be distorted by taxes, tariffs, or subsidies, but nonetheless prices that clear themarket. Now I turn to cases where this is no longer true-cases of licenses, quotas,arbitrary rationing, and the like. To begin with foreign exchange, consider a country witha fixed exchange rate system that uses licensing rather than monetary policy as theprincipal mechanism of adjustment. That is to say, export proceeds must be turned overto the exchange authority, which then doles them out among the many applicants. Ifthere is enough foreign exchange to fill all applications at the going exchange rate, thelicensing system is redundant and the problem reverts to the cases discussed earlier. Butif-as has actually been the case in many countries for long periods-there are far fewerdollars available than are desired, a problem of allocation appears, which the licensingauthorities somehow resolve. Characteristically in such cases, the demands for some"essential" imports will be fully met, others not so fully, others partially, and some (theprohibited list) not at all. It is plausible to assume that the tariff structure probablyalready recognizes this hierarchy, with perhaps zero duties on "essentials," moderateduties on the next group, and progressively higher ones on the remaining two.

What do the three postulates tell us here? Only for the first group would demandprice and supply price be equal to market price. For the other groups the demand pricewould be above the supply price even if all the desired dollars were made available.Because dollars are not made available, there must be some unsatisfied demanderswilling to pay prices well above the world market price plus the tariff. If such demanders

34 Arnold Harberger

are given incremental dollars, their demand prices should be treated as a measure of thebenefit (postulate 1).

The easiest case to deal with is one having an open market for the domestic resale ofthe licensed items. In this case the appropriate demand price is the one prevailing in thatmarket. The social opportunity cost of foreign exchange will be a weighted average of themarket prices of the various categories of imports. The problem is that the weights hereare not derived from the structure of demand and supply and from the normal elasticityof response of different items within that structure to changes in the degree of ease orpressure in the foreign exchange market. Instead, the weights stem from the policydecisions of the licensing authorities, which-particularly with respect to items not in the"essential" category-tend to change with great frequency; items "serve sentences" onthe prohibited list only until enough political pressure takes them off. In a case whereimported goods can be freely resold in the local market, it would be easy to estimate thesocial opportunity cost of foreign exchange if there is a stable, consistent, and predictablelicensing policy. I know of no case, however, where these conditions have been met. Ifpolicy is governed by sufficient rationality to do all these things well, it seems highlylikely that the authorities would then take the next logical step of following monetaryand exchange rate policies that would render the licensing system superfluous.

When there is no open market (either legal or parallel) for the licensed items, theproblem becomes one of estimating their value to the users. The principle of measuringthe social opportunity cost of foreign exchange remains the same: When the licensingauthority gives foreign exchange to a new project, it must come at the expense of otherdemanders. The opportunity cost is then the value (demand price) that those otherdemanders would place on the amounts of which they are being deprived. This isobviously impossible even to try to determine, so the estimation of the social opportunitycost of foreign exchange becomes a crude process indeed. The only solid base to workfrom in this case is that we know that any licensed import is worth at least the worldprice plus the prevailing tariff to all license applicants, because that is what they willhave to pay even if they get the license.

On the side of the social opportunity cost of capital, interest rate ceilings of variouskinds are probably the most frequent source of deviation from a market solution. Thecritical element in these situations is that banks and other financial institutions aresometimes faced with a demand for credit that far exceeds the available funds-especially since the same demanders get in line at many banks. Characteristically, servicecharges, minimum compensating balances, and other devices appear. They may bepowerful enough to produce a market solution-that is, one that involves no arbitraryrationing. In this case one would have to determine the actual total payments by theusers of credit in order to estimate their demand price according to postulate 1, but noother serious complications emerge.

Frequently, however, the equivalent of a market solution is not worked out, andbanks and other financial institutions are left with a substantial range of discretion. Thisopens the door to bribery and corruption, which indeed frequently occur. But there aremore subtle ways, generally fully or nearly within the law, of doling out credit in thesecircumstances. For example, when a borrower hires a banker's relative at a good salary,his subsequent loan applications may more readily be approved. Thousands of ruses ofthis type exist, and given the controlled interest rate, the banker can fairly argue that hischoice among alternative solvent borrowers has cost his stockholders nothing. The net


result is that credit is rationed in inefficient ways, and the marginal productivity ofcapital overall is lower.

To determine the influence of this situation on the social opportunity cost of capital,one must inquire what happens when financial institutions are left with lower lendingcapacity because some savings have been diverted to finance a new project. No clearanswer exists for this case, but the most reasonable assumption is that given the ability offinancial institutions to ration funds, their basic behavior patterns will not changesignificantly. Thus we can say, roughly at least, that in these rationing situations thesocial opportunity cost of the diverted funds will be greater than the controlled interestrate, and less (because of the inefficiency factor noted above) than the typical gross-of-taxproductivity of capital (p in the simplest case) that would prevail in an open capitalmarket.

The above examples show how one can attempt to cope with what I have callednonmarket situations in the measurement of social opportunity cost, using the three basicpostulates as a guide.

The Social Rate of Discount

The social rate of discount has been at the center of controversy for many decades. Earlydebates raged over whether to use the marginal productivity of capital, the marketinterest rate (usually the government bond rate), or the marginal rate of time preference.These concepts correspond broadly to p, i, and r, as presented earlier. The recentdiscussion has been less simplistic. The major contending views, at least as I see them, allrecognize that both p and r, when they are different from i, reflect distortions, and allrecognize that when properly taken into account, this fact leads to situations in whichsocial project evaluations yield different results from private ones. In my opinion, themajor contending views all are basically compatible with the three basic postulates. Thesquabbles are not between saints and sinners, but rather among factions within the samechurch. Their resolution, in consequence, turns on issues of efficiency, relevance,convenience, robustness, communicability, and the like, rather than fundamental error orheresy.

Three approaches to the social discount rate will be treated; they bear a curious sort oftriangular relationship to one another. For example, two of them are based on theconvention that marginal funds come from the capital market, while the third adopts thealternative convention that they come from fiscal sources. In a different pairing, two ofthem use weighted averages of p and r as the discount rate; the third uses r alone.

Consider, first, the dichotomy between the view that the funds for public sectorprojects come from fiscal sources and the alternative view that they come from the capitalmarket. The proponents of fiscal sourcing (Eckstein 1957; Eckstein and Krutilla 1958;Haveman, 1972) argue that most government funds come from fiscal sources and thatmost of the increment in government funding over time has been, and almost inevitablyis, mainly on the fiscal side. In this they are correct. They then proceed to postulate a setof weights, fi and f2, reflecting the fractions of an increment of fiscal revenues that comeat the expense, respectively, of investment and consumption. Their final formula, a socialdiscount rate equal toflp +f2r, is identical in form to the one we derived earlier. The onlydifference is that the weightsfi andf2 here derive from a hypothetical fiscal experimentrather than a hypothetical capital-market experiment.

36 Arnold Harberger

As noted earlier, in a functioning capital market we have reason to expect that thereactions to market pressure or ease will on the whole remain quite similar from one caseto the next. In particular, we can be sure that just about everywhere the elasticity ofinvestment with respect to market pressure substantially exceeds that of saving. Incontrast, experience all over the world suggests that each tax change is very differentfrom the last. The weightsfl andf2 may bound all over the map, depending on whetherone is imposing value added taxes, tightening income tax loopholes, lowering high orprohibitive tariffs to gain more revenue, and so on. To me, it would be perfectly sensibleto have a fiscal weighted-average measure of social opportunity cost if, say, the valueadded tax was the only or principal tax and if changes in the fiscal situation of thegovernments were met by changing its rates. In that world, I, too, would be a proponentof this position.

But the world we observe shows a whole panoply of fiscal adjustments with little orno predictability about what the next one will be like. I find it impossible to conjure upeven a semirealistic mental experiment in which a specified fraction of fiscal revenuestypically comes at the expense of consumption, with its complement coming at theexpense of investment. In contrast, it is natural for these fractions to be relatively stablefor dollars drawn from the capital market. Because of the relative stability of theweighting structure, this argues in favor of the convention that the marginal source is thecapital market.

Of course, a stable weighting structure would not mean much if the idea of the capitalmarket as the marginal source of funds did not make sense. But in fact, on a day-to-day,month-to-month, and even year-end-to-year-end basis, the capital market is in mostcountries the marginal source of funds. Most government budgets, even on the day theyare first presented, contemplate the borrowing of some funds. In addition, as actualevents produce deficits greater than planned, governments almost always turn to thecapital market for the difference. When the future smiles and deficits are smaller thanexpected, the extra money is in effect returned to the capital market. In sum, the capitalmarket is the marginal source when funds are short and the depository for marginalfunds when they are abundant.

Another advantage of adopting the convention of treating the capital market as thesource of funds at the margin is that one can readily adapt it to incorporate capital fundsfrom abroad. the weighted average would now be fAp +f2r +f3MCf, where fi +f2 +f3 = 1and where MCf is the estimated marginal cost of foreign funds. One should employ herethe marginal cost of foreign funds because of the presumption that the supply curve ofsuch funds is not infinitely elastic. With an upward sloping curve, the marginal cost offunds will exceed the average cost.4 This is an obvious consideration in calculating thesocial opportunity cost of capital.

Thus the arguments for building the calculation of the social opportunity cost ofcapital on a capital-market sourcing model are that (1) the weights are relatively stable,(2) the capital market is the de facto marginal source and destination of funds in the shortand middle run, and (3) the calculation is readily adaptable to incorporating sourcingfrom the world capital market.

4. The textbook formula is MC = AC[1+(1/e)], where AC is average cost and E is the priceelasticity of the average cost curve-in this case the upward rising supply curve (of foreign funds)facing the country in question.


Having elected to hold with a capital-market sourcing convention, let us consider therelative merits of using the social opportunity cost of capital (flp +f2r), on the one hand,or the marginal rate of time preference (r), on the other, as the discount rate in social cost-benefit analysis.

The best starting point is to realize that modem defenders of the use of r as the rate ofdiscount (Dasgupta, Marglin, and Sen 1972; Feldstein 1972; Squire and van der Tak 1972)do not neglect the existence of the distortions taken into account in the discount rate byproponents of the weighted-average approach (Baumol 1968; Bruce 1985; Sandmo 1976;Sandmo and Dreze 1971; Sjaastad and Wisecarver 1977). Instead of being reflected in thediscount rate, these same distortions show up in the "shadow price of investable funds"in the case where r is used for discounting. The procedure is as follows. Assume that onedollar is extracted from the capital market. A fraction fl of this comes from displacedinvestment, which in turn would have generated future income at the rate of p per year.The complementary fractionf2 comes from newly stimulated saving, the supply price ofwhich is r per year. There is therefore an annual opportunity cost of fip+f2r for each yearin the future as a consequence of withdrawing one dollar this year from the market. Thisfuture flow of opportunity costs, discounted back to the present at the rate r, has apresent value of (fip +f2r)/r. This is the shadow price of investable funds for those whouse r as the discount rate.

If p = 12 percent and r = 4 percent, withfi = 0.75 andf2 = 0.25, we would have aweighted-average opportunity cost of capital of (0.75)(12 percent) + (0.25)(4 percent) = 10percent. This is what the weighted-average advocates would use for discounting. Indoing so, their investment criterion would be that the discounted value of benefits shouldexceed the discounted value of costs.

In contrast, the advocates of discounting by the marginal rate of time preference (r)would use 4 percent as the discount rate, but would require that the present value ofbenefits be more than 2.5 times the capital costs of the project. The factor 2.5 is exactly(fip +f2r)/r; discounted at 4 percent, this is the present value of what is given up forevery dollar withdrawn from the capital market.

If the two sets of criteria (10 percent discount rate requiring that benefits exceed costs,and 4 percent discount rate requiring that benefits exceed 2.5 times costs) were properlyimplemented, I do not believe there would be many serious contradictions in theirimplications for longer-term investment projects. However, I have never had any doubtabout preferring the first approach. Three grounds for this preference are:communicability of the procedure, implications with respect to current expenditures, andimplications for handling situations with different rates of time preference for variousgroups.

With respect to the first, I have always believed that the most basic function of projectevaluation was to shoot down the worst projects. Unfortunately, most of the worstprojects have strong supporters, usually within and outside government. To my mind,the project evaluation team weakens its position if it adopts a criterion that requiresbenefits to be, say, 2.5 times costs.5 It is hard to beat down a project with the argumentthat its benefits are only twice its costs, and that this is not enough!

5. To be sure, within the context of the methodology, the factor 2.5 is necessary to adequatelyreflect the present value of the future costs entailed in borrowing. Once this factor is used as ashadow price to multiply capital costs, the correct rule is that benefits should simply exceed costs,

38 Arnold Harberger

With respect to the second point, I have always been greatly impressed by a powerfulargument raised by Sjaastad and Wisecarver (1977) in a paper that merits the mostserious attention. Once the capital market is accepted as the marginal source of funds, itshould be recognized as such not only for capital outlays, but also for current spending.Most particularly, any money saved through greater efficiency in the police force or inthe schools can be used to pay off debt; used in this way, it will (in our numericalexample) produce a benefit equal to 2.5 times its nominal amount. The logic of the casepresses one ineluctably to the conclusion that for current outlays as well, the critical ratioof benefits to costs should be 2.5. In short, the factor (fip +fhr)/r represents the socialvalue of one dollar of liquid funds, either taken from or placed in the capital market. Anyfunds sourced in the capital market will have a shadow price of 2.5 per dollar, in ournumerical example. Also since the capital market is being taken as the marginal source offunds (as in truth it really is, in most cases), the 2.5 factor should apply to all cash costsacross the board. Cutting out any cash outlays-either current or capital-wouldlikewise permit achievement of benefits at the ratio of 2.5 to 1. A powerful andpersuasive argument, to which I have yet to see a convincing rebuttal!

The third reason for preferring the weighted-average discount rate to the marginalrate of time preference is my own (in the sense of my not having encountered it in otherwritings or discussions). To my mind, the weighted-average discount rate movesnaturally from the simple fip +f2r to the more disaggregated .gfj + £kfk rk. In each casethe p1 are the demand prices of displaced investments and the rk are the supply prices ofnew(y stimulated savings. A project undertaken today must be able to repay the cost ofits assets and to cover these demand and supply prices; otherwise it has not paid its way.There is nothing in the weighted-average approach that insists there be only one r, andthe existence of any number of different rk presents no conceptual difficulty orembarrassment.

Contrast this with the use of r as the discount rate. Its justification stems from thestandard treatment of intertemporal consumption decisions in economic theory-the rateof return to the saver, after all taxes, is the rate that he uses in deciding on his savings.The weighted-average approach builds on this also, but the time preference advocatesturn this into the cornerstone of their discounting procedure, on the ground that it is byusing r, not p or i, that consumers make their intertemporal choices. Here is where Iencounter difficulties in passing from a case in which there is a single r (a single uniformtax rate separating i and r for all consumers) to one in which there are two or more rk fordifferent groups (for example, two or more marginal tax rates). If r is used as the discountrate because it is the fundamental guidepost for consumers as they make theirintertemporal consumption decisions, then when various consumer groups use differentrates, it would seem that each group's benefits should be discounted at its own rate.

To my knowledge, none of the advocates of using r as the discount rate has broachedthis problem-I merely say it is a problem that flows naturally from the conceptual basisof the time preference approach. The underlying problem is that when there are differentrates of discount for various groups, postulate 3 needs tighter specification. Suppose

so adjusted. But nonprofessionals will still wonder about projects being rejected simply because thepresent value of their benefits is only 1.5 or 2 times actual capital outlays, and powerful forcesbehind bad projects would surely use such circumstances to convince the public and the relevantauthorities that the projects were in fact quite good.


groups A and B have different rk's, and suppose there are two projects (say, parks), eachof which gives to one of the groups a certain path of in-kind benefits through time. Letthese benefit paths be identical for the two groups. If we simply subtract one benefitprofile from the other, year by year, the difference is zero each year. But if we discountfuture benefits back to the present, the group with the lower discount rate (presumablythe wealthier of the two) will have a higher present value of benefits. Also, if weaccumulate the benefits forward to some future time, measured as of that point in time,the group with the higher discount rate will have the higher present value of benefits. Soby subtracting the benefits of A from the benefits of B year by year, we perceive nodifference; by discounting to the present, we find group A with greater benefits; byaccumulating to the future we find group B with greater benefits. (There seems to be nogood reason to treat costs differently in the two cases, as they are supposed to followidentical time paths.) All this runs counter to the standard proposition of capital theoryand project evaluation that the ranking of projects should not depend on the point intime to which benefits and costs are discounted or accumulated; it matters only that theprojects be evaluated as of the same point in time. This proposition is always true if asingle rate is used for discounting and accumulating- even when that rate varies fromone year to the next. But it obviously is not true with rates applying to groups ofeconomic agents.

Actually, those who propose using time preference discount rates always work with asingle rate, which avoids the ambiguity just referred to. I concur and suggest that insocial project evaluation, postulate 3 should apply within the year, canceling costsagainst benefits as they occur, and that a single discount rate (possibly varying from yearto year) should be used to carry net benefits or costs from one time period to the next. Tomy mind, the weighted-average social opportunity cost of capital (Zfjpj + Ek fk rk ) isideally suited to this role. For the reasons already expounded, a time preference rate isnot. 6

Concepts versus Numbers

Practitioners whose main experience is in the field, together with others who functiondaily in the world of affairs, may wonder at the level of precision that has characterizedmost of the discussion so far: demand prices, supply prices, distortions, and weightsbased on elasticities that are impossible to observe directly and virtually impossible toestimate exhaustively (that is, it is impossible to estimate all the items necessary toconstruct a system of elasticity-based weights). We have weights that depend on savingand investment behavior, for which econometric explanation is still a matter ofcontroversy even in advanced countries. We have weights that depend on theresponsiveness of individual classes of exports and imports to the tightening and easing

6. 1 am assuming that there are different groups with different values of rk. The weighted-average social opportunity cost has a precise meaning-the amount needed to compensate alllosers when money is drawn from the capital market. Using as the discount rate a weightedaverage of just the rk's will do the trick in a mechanical sense-that is, it will eliminate theambiguity of working with separate rk's-but no clear concept lies behind such a weighted averagecovering only marginal rates of time preference.

40 Arnold Harberger

of general pressure on the exchange market, items that to my knowledge have not beenestimated at all. 7

What then do people actually do? In point of fact, most of the time they use extremelycrude estimates based on readily available data. The simplest measure of the socialopportunity cost of foreign exchange is the market exchange rate augmented by the so-called force of tariff-the ratio of actual tariff receipts to the c.i.f. value of imports. Thismeasure (1) neglects the export side altogether, (2) makes no attempt to distinguishbetween import categories that are especially sensitive and those that are especiallyinsensitive to changes in the real exchange rate, and (3) completely ignores the possibilitythat imnportant distortions (such as petroleum and gasoline taxes) that do not fall directlyon trade may nonetheless have a measurable effect on the social opportunity cost offoreign exchange.

The great benefit of having a conceptual framework is not only that it helps us thinkthrough problems in a clear way. It also tells us how to try to improve on estimates andmeasures that are extremely crude and approximate. Starting with a force-of-tariff firstapproximation, the steps we must take are those suggested by points 1-3 above. First, wemust ask whether exports belong in the weighted average. If the exchange policy is oneof simply doling out by licenses whatever foreign exchange comes in, it may well be thatno export adjustment should be made. But if market forces determine the real exchangerate to a substantial degree, then one should try to infer how elastic is the response toexchange rate pressure of different categories of exports, and to determine whatdistortions (such as taxes and subsidies) are relevant for each major category. At the veryleast, a broad "average distortion" for all exports taken together (for example, the ratio ofexport taxes or subsidies to their total f.o.b. value) can be estimated and introduced alongwith the force of tariff to produce a weighted-average estimate (including exports) of thesocial opportunity cost of foreign exchange.

Second, to break down imports into categories that have different relative sensitivitiesto changes in the real exchange rate, the practical procedure is to start, with simply avalue-weighted average of all tariffs. (If exports are not in the picture this actually yieldsthe force-of-tariff ratio.) Then one classifies the categories into those that are judged to be,say, very sensitive, somewhat sensitive, average, somewhat insensitive, or veryinsensitive to changes in the real exchange rate. Having done so, one adjusts the weightsaccordingly-doubling or tripling, say, the weight attached to the very sensitive groupand perhaps cutting by one-half or two-thirds the weight attached to the most insensitive

7. In deriving the relevant weights in the foreign exchange case, the origin of the pressure isconceived of as additional purchases of foreign exchange in the market. These purchases raise thereal exchange rate either because of the flexibility of the nominal rate or because of the naturalworkings of the adjustment process under a fixed exchange rate. If the real exchange rate isdenominated by e, the derivatives that are relevant for constructing the weights fl, and fk areM1 /be and 8Xk/&e. For example, fl would be equal to -( &4/&)4-X (I4/&)/ + 1* (4k/e)] and fkwould be &Xk&/& divided by the same denominator. 6M1 /& is typically negative when e is defined,as it is here, as the local-currency price of, say, the dollar.) In short, we are asking how each item inthe foreign exchange market responds to the general pressure of a tightening of the real exchangerate. This is not the same thing as how each item responds to a change in its own price or to anincrease in pressure in just its own market.


group. Any category subject to a quota that is unlikely to change but likely to becontinually effective would receive a zero weight, because these imports, even thoughthey may be quite important, are neither displaced when the real exchange rate tightens,nor do they expand when it eases. A similar treatment can then be applied to exports.Obviously, throughout the process of adjusting weights, take care not to alter their sum.

The third step of adjusting the social opportunity cost of foreign exchange for internaldistortions is relatively easy. If, say, the country produces no petroleum, then thepetroleum and gasoline taxes imposed internally function in the same way as tariffsurcharges-one simply combines them with the tariff to get a picture of the totalgovernment revenue generated by each dollar of petroleum imports. In case there isdomestic production of a good (such as alcoholic beverages) that is subject to excise tax,an increase in the real exchange rate will give rise to some substitution toward thedomestic product, hence the government's loss in excise taxes will be some fraction ofwhat one would predict on the basis of the reduction in imports.

The basic conceptual framework also guides us as we struggle to come up withempirical estimates of the social opportunity cost of capital. The fact that the concepts areframed in real terms suggests beginning with direct, rather than indirect, measures of thereal return to capital. Rates of total return, including real capital gains, in financialmarkets (especially stock markets) are notoriously volatile and subject to the vagaries oftransitory swings in expectations. Much more steady and reliable are measures of thenational or sectoral rate of return, based on estimates of the real capital stock (usuallybuilt up via a perpetual-inventory process) together with data on the real returnsaccruing each year to the capital factor. Considerations of differential taxation suggestdistinguishing where possible between predominantly corporate and predominantlynoncorporate sectors, and singling out those with special tax treatment, such as housing,mining (at times), and agriculture (frequently).

The underlying concept is one in which the government draws its marginal fundsfrom the capital market. In most countries, this probably means that there is nosystematic mechanism by which one government project derives part of its funds bydisplacing other investments throughout the public sector. One project may indeeddisplace another-even in its entirety-but such displacement is likely to be unique foreach project, not systematic and similar for all classes of public projects.

Consequently, when we try to estimate p, the marginal productivity of displacedinvestment, we probably should try to measure what I call the social rate of return toprivate sector capital. We can look at the case where government project A in factdisplaces government project B then as a combination of two modules: one in whichproject A is undertaken and is financed by resort to the capital market; the other in whichproject B is canceled and its funds returned to the capital market.

Some countries have only rudimentary capital markets, often pretty much limited tothe banking system itself. In such cases, government's resort to the capital market usuallymeans that a roughly equivalent amount of credit is crowded out of the private sector.Therefore the first approximation of the marginal productivity of displaced investmentshould be the real social rate of return to capital by those segments of the private sectorthat are the typical recipients of bank credit. Where monetary policy does not permitgovernment borrowing from the banking system to crowd out enough investment, thetypical result is that part or all of the government's borrowing gets reflected in inflation.It is then likely that a significant part of the government's outlays will in the end come at

42 Arnold Harberger

the expense of consumption. Inflation processes are exceedingly complex, however, withinstitutional arrangements (such as indexing) varying widely among countries. In aninflationary environment, therefore, analysts should probably treat the weights such asflandfk as a problem to be solved separately for each country.

With respect to the labor market-perhaps the most vexing case of all-theunderlying concepts are essential guideposts to those who undertake the difficult task ofempirical estimation. From the very outset, the concepts tell us to seek the ultimatepattern of sourcing from which an increment to labor demand is filled-and not to taketoo seriously the provenance of the particular workers hired by a new project. They maybe lured away fromn other employers; they may quit a job because the project is near theirhome or otherwise appealing; or they may be picked up by the project during atemporary spell of unemployment. This direct sourcing pattern may be next tomeaningless: If they left another job, the likelihood is they will be replaced; if they wereunemployed, they would probably soon have found some other job.

At the other extreme, the concepts warn us away from a demographic-historicalapproach that considers the ultimate sources of new labor supply to be natural increase,migration into the region, and changes in the rate of labor force participation. Theproblem with this approach is that most of such changes would have taken placeregardless of the presence or the absence of any particular project. What we are seeking isthe chain of causation by which the presence of a particular project draws labor (directlyand indirectly) from where it would have been (in equilibrium) in the absence of theproject. In short, we must think about sourcing in economic rather than demographic-historical terms.

To think in economic terms means to think about markets-in this case the market forthe particular type and class of labor involved. One must determine whether the marketis a national or a regional one and, if the latter, to what degree migration will likely meetincrements of demand. One must try to identify the structure of the market and thecharacteristics of its supply and demand. For highly specialized and highly skilled jobs,the likelihood is that the bulk of any new demand will be met by bidding workers awayfrom other employers. For relatively unskilled, undifferentiated tasks, the existing pool ofthose who hold such jobs has little significance-new taxi drivers, hotel maids, or grocery

clerks can come from anywhere.By fortunate coincidence, dealing with the social opportunity cost of labor is

simplified by the fact that most labor market distortions (such as payroll taxes andincome taxes) are relatively modest in size (at least compared with many tariffs and withsome taxes on the income from capital) and also rather widespread through the laborforce. Perhaps the biggest distortions in developing countries are those between the mosthighly protected sectors (typically multinationals) and the rest of the economy. But thiscreates relatively little difficulty because we can be sure that the other sectors do notobtain any net labor from the most protected ones. If a worker leaves a highly protectedsector, a long list of candidates is always waiting to replace him. The highly protectedsector's wage enters the calculation of social opportunity cost only when the new jobsbeing created are actually in that sector. In such a case the protected-sector wage entersas the upper limit to the supply price of the quasi-voluntary unemployed.

Similarly, although the problem of determining empirically the social opportunitycost of the unemployed in a cyclical setting poses challenging conundrums to the analyst,


it is of relatively minor importance in real-world project evaluations. 8 The reason is that,most of the time, project evaluations are forward-looking operations. The project beingevaluated may not even start until two or three years later, and its economic life maythen stretch for ten or twenty or even fifty years. Whether or not cyclical unemploymentis currently observed, the best prediction for future years is that the situation will be"normal" or "average." As far as I can see, the issue of the social opportunity cost of laborin a setting of cyclical unemployment comes up largely with respect to programs(perhaps of temporary job creation) specifically designed to deal with the unemploymentsituation. In this context, there will be a relatively rapid absorption of the unemployed,with a duration roughly commensurate to the recession under way. The typical longer-term investment project simply does not fit into this context.

With respect to chronic unemployment, the two varieties discussed were: quasi-voluntary unemployment linked to the presence of protected sectors, and migration-fedunemployment. Neither of these, in my view, presents conceptual or measurementproblems of a serious nature.

Distributional Weights and Basic Needs

At the outset I want to re-emphasize that all three basic postulates form the roots of thegrand tradition of applied welfare economics; all three are needed to produce the seriesof classic results recounted at the beginning of this paper; all three have been employedby a long lineage of great economists. Nonetheless, it is the third postulate that comesinto question most often and has raised doubts among some thoughtful people-economists and noneconomists alike. The purpose of this section is to address suchdoubts and to discuss some of the issues surrounding them.

Let me begin at the end. The great economists who have employed the threepostulates have been neither naive nor crass. From their writings one can readily discernthat they do not believe that the postulates and their implications are all that count. Eachsociety has its own values, and each has many important objectives apart from economicefficiency. No one, to my knowledge, has argued that the three postulates shouldoverthrow society's values or should supplant important noneconomic objectives. But thethree postulates do provide a disciplined, coherent framework for thinking through theeconomic aspects of a wide range of problems. The appropriate analogy is withaccounting-another example of a complex structure of analysis and consequences thatrests ultimately on a few fundamental postulates. Accounting tells how to calculate netprofit and net worth (wealth). It does not attempt to do the absurd by saying that profitsand wealth should never be compromised or sacrificed for other objectives. Accountantsknow and recognize this. So do economists when it comes to considering theconsequences and implications of the three basic postulates in comparison with otherobjectives.

S. The principal problem here is to determine what supply price of labor to use. Labor supply,particularly of adult males, is known to be extremely inelastic in most countries, yet in repeatedsurveys the unemployed state their supply price close to the going wage for their age-education-experience category in their own labor market; that is, the stated supply price takes on the attributeof the expected demand price in the market. The question is: To what extent should we accept suchstatements of supply price at face value, thus attributing little gain in utility to the unemployedperson when he or she finds a job?

44 Arnold Harberger

One must be careful here, however. Traditional accounting, although far fromrevealing the meaning of life, is important enough in many different contexts to be takenseriously. Also part of accountants' role is to call the attention of others to the relevanceand usefulness of their discipline. So, it seems to me, it is (or should be) with appliedwelfare economics. It is significant and relevant for many matters of interest to society. Itgives a disciplined and structured response to important questions. It embodies some ofsociety's values, but not all. And in many, even most, cases its results are notincompatible with society's other values. In my view, part of the role of economists is topresent the answers given by applied welfare economics to the problems at hand, as theparticular contribution of the economics profession to society's decisionmaking process.If economists do not present what economics as a discipline has to contribute, who elsewill?

The approach just described is modest, in the sense that it makes no excessive claimsfor the postulates and does not try to graft onto them other objectives that might expandtheir scope. A significant strand of literature, however, follows a different line,specifically with respect to the distribution of benefits and costs. This literature exploresthe use, within applied welfare economics, of distributional weights. Use of such weightsentails multiplying the net gains or net losses of particular groups by specific factors-higher for groups whose welfare is deemed more meritorious, lower for those whosewelfare is less prized (presumably by society as a whole). In principle, welfare weightscould be classified according to any criteria-ethnic or national origin (Malaysians andChinese in Malaysia), educational or class background (untouchables in India), type ofeconomic activity (independent farmers and farm laborers in many countries), and so on.But most of the literature has focused on discussing income or wealth as the criterion andhas typically assumed that the welfare weights are a declining function of the criterionvariable.

Several authors have rigorously applied such declining weights, particularly to dealwith the problem of the optimal choice of an income tax schedule (Atkinson 1970;Atkinson and Stiglitz 1976, 1980; Diamond and Mirrlees 1970; Dixit and Sandmo 1977;Mirrlees 1971, 1979). A few attempts have been made to incorporate distributionalweights in actual project evaluations. Most of these have been flawed by the applicationof the weights to the changes in income of the various groups, rather than to the changesin surplus (net benefit or cost). All the logic of distributional weights leads to the use ofchange in consumer and producer surplus as the base to which to apply thedistributional weight. An increase in labor along an infinitely elastic supply curve, forexample, implies no increase in welfare to the suppliers; hence there is no relevant baseto which to apply the group's distributional weight.

After giving considerable thought to the problem, over almost a decade, I have cometo the conclusion that systems of distributional weights do not adequately represent theway citizens feel about redistribution efforts channeled through the public sector.Instead, I believe that a system of "basic needs externalities" reflects people's values andbeliefs much better than distributional weights (Harberger 1984). Basic needsexternalities are based on the idea that citizens and taxpayers look for specific andconcrete results when public funds are channeled into helping others. They want to seethe recipients turn out to be better educated, better cared for medically, better fed, andbetter housed. In short, citizens and taxpayers are not interested in having their moneyused simply to gratify the recipients; they want to see it used to advance the welfare of


the recipients as they (the donors) perceive that welfare. This helps explain why the mostuniversally popular government transfer programs have entailed transfers in kind ratherthan in cash, as in the case of universal free primary education and free or heavilysubsidized medical care for the indigent. It also explains why public support has beenweak for programs in which transfers in kind can be converted into cash with relativeease. Subsidized food, or the food stamps with which to acquire it, can be resold;subsidized housing, with space judged to be "adequate," may quickly becomeovercrowded by an inundation of "cousins." In such cases the recipients are maximizingtheir own welfare by transforming their in-kind transfer into cash, but the donors areunhappy because their purpose (improving the welfare of recipients as the donorsperceive it) has not been served.

This and much other behavior, as well as many other opinions and attitudes on thepart of citizens and taxpayers, can be rationalized in terms of basic needs externalities.Citizens, the argument goes, are concerned enough to have their money spent to ensurebetter education, medical care, nutrition, and housing for the less fortunate members ofsociety. But donors are not interested in seeing their money go toward financing moreelegant dowries for brides, more elaborate funeral ceremonies for grandparents, or morefrequent return trips of workers to their native villages. Even if these are the things therecipients prefer, they give rise to no external benefit as perceived by the donor, hencedonors (taxpayers) do not want to see their money used directly or filtered indirectly topay for such items.

Elsewhere I have suggested ways of getting crude quantifications of the basic needsexternality for such needs as education, medical care, and nutrition (Harberger 1984).These quantifications are justified by the existence of a basic needs externality, but theyare also influenced by a modified version of the least-cost principle. The least-costprinciple tells us not to accept any project (or other action) if a cheaper way of achievingthe same or equivalent benefits can be found. This dictum needs to be modified, at leastfor basic needs externalities, because in most countries an almost infinite number ofsituations exist in which equivalent benefits from basic needs externalities might befound. The modified least-cost principle, then, states that the costs society is willing toincur to meet a basic need should vary inversely with the intensity or urgency of thatneed. For example, we would typically be willing to incur a greater cost to bring a givengroup from 85 to 90 percent of some nutritional or health standard, than to bring asimilar group from 95 to 100 percent of that standard. Similarly, poor countries canattempt to subsidize only the earlier stages of education for broad groups of thepopulation; the higher the level, the more selective the criteria must be.

In line with this modification principle, the project authority could set, for each basicneed, a cutoff level above which no basic needs externality is deemed to exist. Morever,since the externality in the end turns out to be a justification for society to pay some or allof the bill, it is important that the income or standard of living of the recipienthouseholds enter into defining the cutoff level. For most developing countries, forexample, it does not make sense for society to subsidize the diets of the top 50 or 60percent of the population. Cutoff levels for the attribution of any basic needs externalitywould presumably vary with the type of need. They might be set at the 75th percentile ofthe income distribution for primary education, at the 40th for secondary education, andat the 25th for higher education. For basic medical care, they might be at the 60thpercentile, while for housing only at the 20th. For society to attribute a basic needs

46 Arnold Harberger

externality, in the terms in which we are speaking, it must be willing and able to bearpart or all of the cost.

In addition to establishing a cutoff point for each externality, the modified least- costprinciple would define a maximum externality (as a percentage of the normal cost ofproviding the service in question) that society is willing to attribute, even in urgent cases.The reason is similar to the one that motivated the original version of the least-costprinciple. If we attribute a maximum externality of 100 percent of normal costs in cases ofvery low nutrition levels, we can be sure that, as a result, society will pay double thenormal cost at least some of the time to meet the need in question. The logic of cost-benefit analysis is simple and ineluctable. If the normal cost of meeting a given (say,nutritional) need is 10 per unit, and if we attribute a 100 percent externality to meetingthat basic need, in two plausible extreme cases that externality would cause theacceptance of a project that would otherwise be rejected. At one extreme is the case inwhich there are no benefits of the usual type-the externality is the only benefit. In such acase the project would be accepted if its costs were less than or equal to the usual 10 perunit. At the other extreme is the case in which the usual type of benefit is equal to 10 perunit; here the existence of the externality will render a project acceptable if costs range upto 20 per unit. In both cases society ends up paying a cost of up to 10 per unit as the"price" for meeting the basic need in question.

Could not cheaper ways of meeting this or an equivalent basic need be found? Almostcertainly so, and the surest way to bring about a search is to disallow the attribution ofexternalities as large as 100 percent in the first place. When we attribute a givenexternality as a benefit, we say in effect that it is worthwhile to incur a correspondingamount of costs in order to produce that benefit. Without a systematic process ofsearching for alternative, cheaper ways of obtaining given amounts of benefits, theattribution of an externality is an invitation to incur up to that amount of extra costs so asto achieve the external effect.

The modified least-cost principle embodies the idea that society is willing to acceptthe risk of inefficiency in the most urgent cases, such as famine, starvation, andpestilence. But the willingness to accept this risk declines as the basic need becomes lessurgent, until-at the cutoff point-society simply refuses to attribute any externality andthereby refuses to invite the conscious acceptance of extra or higher costs.

The practical implementation of the schema just described can be made very simple.The budget bureau, the finance or planning ministry, the project evaluation office, oreven the national cabinet would decide that for, say, the lowest percentile of householdsit would assign a basic needs extemality of 30 percent of the normal cost of additionalnutrients. This percentage would decline to zero at the 40th percentile of households (thecutoff point). Thirty percent, in this case the maximum allowable externality, is also themaximum amount of excess cost that the procedure would permit. Society would beaccepting some inefficiency (by standard criteria) for meeting basic needs. But thisacceptance would be tempered by placing explicit and conscious limits on the extra coststo be incurred on this account.

Although the use of distributional weights by definition entails the rejection ofpostulate 3, the attribution of basic needs externalities is fully compatible with it. Thepositive extemality involved in improving the education, health, nutrition, and housingof the disadvantaged takes its place alongside other extemalities, such as air and waterpollution and traffic congestion. All these, like basic needs externalities, have the


attributes of public goods (or public "bads"), but all can easily fit as positive or negativedistortions into the framework of TiDiAXi.

Once this is recognized, it helps solve what many people consider a troublesomeconundrum. As shown above, when the relevant group has an infinitely elastic supply oflabor, no distributional weight benefit can be attributed, even when the employment oflow-income workers expands dramatically at the given supply price. This bothers manypeople, who firmly believe there should be a social gain. The answer is clear: The peoplewho are troubled by the no-benefit result do not think in terms of distributional weights.Their intuition, it seems to me, runs more along the lines of basic needs externalities. Arise in employment-even along an infinitely elastic curve, and either through increasesin numbers employed or in hours worked per person-will increase the cash income ofthe affected families. These added funds, in turn, will almost certainly be spent inmeeting at least some additional basic needs. When we value these increments of welfareas externalities using EjDiAXi, we find that the added employment did indeed bring asocial benefit. The basic needs approach thus solves the conundrum.

Epilogue

I have tried in the preceding pages to outline an approach to social project evaluation thatis conceptually sound, simple, and deeply rooted in the grand tradition of economicscience. This approach depicts what economic science has to say or to offer in the area ofcost-benefit analysis as society struggles with the many difficult choices it faces.

In the process I have recognized several times that it is an austere analytical structurethat emerges when we build on the three basic postulates. No more than the accountant'srules does that structure reveal the fundamental purpose and meaning of life or theultimate values that a particular society ought to treasure or to seek.

These things are the source of vast amounts of misunderstanding and confusion.Neither economics as a science in general nor the three basic postulates in particular tellus that individuals and societies should seek to maximize their incomes. Economicspostulates that individuals and families seek to maximize their own welfare (as theyperceive it), including the various ways-such as basic needs externalities-in which thewelfare of others impinges on their own. This welfare-seeking behavior by eachindividual economic unit forms the foundation stone of the three basic postulates. These,in turn, give rise to an analytical structure that extends to society as a whole a procedurefor weighing social benefits against costs, one that is compatible with welfare seeking atthe level of the individual unit.

The resulting structure is austere. It does not answer all questions. It does notincorporate all society's values. It does, however, incorporate the value of economicefficiency in the sense of trying to maximize perceived welfare-not income. Oneimportant question for doubters to ask themselves is: How many of society's other goalsand values are really antagonistic to economic efficiency? I, and the whole tradition ofeconomic science, see relatively little problem of incompatibility. But there may be someincompatibility nonetheless, as when a society chooses to favor certain groups (such asdisadvantaged minorities) in ways that may not fit neatly as basic needs externalitiesunder the three postulates.

Even in such cases economic science can help us reach an answer. For at the very coreof economics is the notion that when we are striving after two or more good ends, but

48 Arnold Harberger

reality tells us that we cannot have more of all of them, we should be prepared tosacrifice something of one to get more of another. This is the most fundamental ofeconomic principles. It tells us that in the presence of multiple social goals we should firststrive to seek compatibility between those goals and economic efficiency; but if thereremains some incompatibility, we should be prepared to make at least a minimal sacrificeof economic goals to gain advances toward other important ends.

But I still remain a defender and champion of the three postulates and of thetraditions of economic science that they represent. So far, I have mainly used the sameterms as the philosophical doubters. In their discussions, the other values that vie witheconomic efficiency are somehow deemed almost always to be good values, with highethical and moral connotations. Almost always they are values that carry us to a worldthat is somehow better than the one we reach using the three basic postulates.

What I see, however, and what I believe was seen by most representatives of the greattradition of applied welfare economics in the past, is a world in which good and evil existside by side, and pervasively. The noneconomic goals pursued in the real world are notall based on the high principles of ethics and philosophy. Many, far too many, reflect thedarker side of human nature.

One important noneconomic goal in most countries-at the very least it is a goal towhich the three postulates have been able to contribute only little-is national defense.This is a perfect example of what I mean, for under the label of national defense we haveseen things that run the full gamut from the most honorable to the most vile. Those whostruggle nobly against aggression and oppression argue their cause in terms of the valueof national defense. Those who perpetrate these very acts of aggression argue their owncause by using the same vocabulary.

Almost by its very nature, social project evaluation is enmeshed in the structure andprocesses of government. Most often it is public sector projects being analyzed-theanalysis done by one public agency, the approval of another required, and theimplementation done by a third.

When we speak then of noneconomic goals, let us not forget the many projects thathave been carried out just to satisfy the caprice or whim of some powerful figure or clan.Let us not forget the corruption that pervades the decisionmaking and contractingprocess in many parts of the world. Furthermore, even when these elements are notpresent, let us not forget how project choice gets intertwined with the political processalmost everywhere: How the granting and withholding of projects are used to rewardpolitical supporters and to punish enemies, and how in electoral situations governmentstend to distribute projects with the purpose of winning over constituencies that may bedoubtful or wavering.

All kinds of objectives compete with the three postulates, and I have no doubt that, inthe final analysis, the less than noble motives enter with greater frequency andimportance than the lofty ones. Viewed in this light, the three basic postulates provide away of insulating the methodology of social project evaluation from the banal, crass,even vile pressures just alluded to. They lead to the sort of professionalism discussed inthe introduction to this essay. Just like the principles of accounting, the methodologybased on the three postulates enables one group of project evaluators to review or"audit" the world of another. The methodology itself should always lead to the sameanswers; the only serious problems lie in estimating and quantifying future costs andbenefits. This aspect is what makes it possible to think of a "professionalized" discipline


of social project evaluation in spite of the many human failings that characterize theenvironment of making and executing project decisions. I believe, too, that thedevelopment of a rigorous professionalized discipline provides the best hope forimproving actual performance and gradually overcoming the institutional weaknessesand personal temptations that have blemished records from the past.

References

Atkinson, Anthony B. 1970. "On the Measurement of Inequality." Journal of Economic Theory 2(September): 244-63.

, and Joseph E. Stiglitz. 1976. "The Design of Tax Structure: Direct Versus Indirect Taxation."Journal of Public Economics 6: 55-75.

. 1980. Lectures on Public Economics. New York: McGraw-Hill.

Baumol, William. 1968. "On Weighted Average Measures of the Social Opportunity Cost ofCapital." Discussion paper, University of California, Los Angeles.

Dasgupta, Partha, Stephen A. Marglin, and Amartya Sen. 1972. Guidelines for Project Evaluation.New York: United Nations Industrial Development Organization.

Diamond, Peter A., and J. A. Mirrlees. 1970. "Optimal Taxation and Public Production: Tax Rules.:American Economic Review 61 (March and June): 8-27 and 261-78.

Dixit, A. K., and Agnar Sandmo. 1977. "Some Simplified Formulae for Optimal Income Taxation."Scandinavian Journal of Economics 79: 417-23.

Dupuit, J.A. 1952. "De la Mesure de l'Utilite des Travaux Publics," reprinted in InternationalEconomic Papers 2: 83-110. London: Macmillan.

Eckstein, Otto, and John V. Krutilla. 1958. Multiple Purpose River Development. Baltimore, Maryland:Johns Hopkins University Press.

Feldstein, Martin. 1972. "The Inadequacy of Weighted Discount Rates." In R. Layard, ed., Cost-Benefit Analysis. Harmondsworth, England: Penguin.

Harberger, Arnold C. 1971. "On Measuring the Social Opportunity Cost of Labour." InternationalLabour Review 103(6): 559-79.

. 1980. "Vignettes in the World Capital Market." American Economic Review 70(2) (May): 331-38.

. 1984. "Basic Needs Versus Distributional Weights in Social Cost-Benefit Analysis."Economic Development and Cultural Change 32 (April): 455-74.

Harris, J. R., and M. P. Todaro. 1970. "Migration, Unemployment and Development: A Two-SectorAnalysis." American Economic Review 60 (March).

Hotelling, Harold. 1983. "The General Welfare in Relation to Problems of Taxation and of Railwayand Utility Rates." Econometrica 6(3).

Lerner, Abba P. 1936. "Symmetry of Export and Import Taxes." Economica 3(11): 306-13.

Lewis, W. Arthur. 1984. "Development Economics in the 1950s." In Gerald M. Meier and DudleySeers, eds., Pioneers in Development. New York: Oxford University Press.

Marshall, Alfred. 1890. Principles of Economics. London: Macmillan.

Mirrlees, J. A. 1971. "An Exploration in the Theory of Optimum Income Taxation." Review ofEconomic Studies 38 (April): 175-208.

-. 1979. "The Theory of Optimal Taxation." In K. J. Arrow and M. D. Intriligator, eds.,Handbook of Mathematical Economics. Amsterdam: North-Holland.

50 Arnold Harberger

Pigou, A. C. 1932. The Economics of Welfare, 4th ed. London: Macmillan.

Ramsey, Frank P. 1927. "A Contribution to the Theory of Taxation. Economic Journal 37 (March): 47-61.

Sandmo, Agnar. 1976 "Optimal Taxation-An Introduction to the Literature." Journal of PublicEconomics 6 January): 37-54.

, and Jacques Dreze. 1971. "Discount Rates for Public Investment in Closed and OpenEconomies." Economica 38(152) (November): 395-412.

Sjaastad, Larry A., and Daniel Wisecarver. 1977. "The Social Cost of Public Finance.: Journal ofPolitical Economy 85 (June): 513-47.

Squire, Lyn, and Herman van der Tak. 1972. Economic Analysis of Projects. Baltimore, Maryland:Johns Hopkins University Press.

4

Project Evaluationfor the Next Decade

Arnold Harberger

As the Economic Development Institute (EDI) strives, once again, to give greater weightto project evaluation among its many activities, it is only appropriate to take a fresh lookat the subject to identify those aspects likely to be most relevant and important in theyears irunediately ahead.

In my view, a lot has happened since the field of project evaluation first came to fullflower in the 1960s. Most importantly, the world has changed dramatically, both in termsof the types of policies pursued and in terms of the way in which economic policymakerstend to define their tasks.

At the time of the blooming of project evaluation as a field (or subdiscipline) ofeconomics, the most common attitude of developing countries was that they had to plan,guide, and direct their economies. Development planning was the watchword;input-output matrices provided the data needed for wise, long-run decisionmaking;tariffs, quotas, prohibitions, license requirements, multiple processes of authorizationand approval-these were the tools by which the authorities sought to channel, guide,and direct development of their countries.

The consequence of this type of attitude and this sort of approach to the developmentprocess was a situation in which it was commnon for market prices to be greatly distorted.That was a world in which an import tariff of 40 or 50 percent was thought of as low,while tariffs of 80 and 100 and even 200 percent were common. On top of tariffs,countries maintained long lists of prohibited imports-lists that, much to theconsternation of project evaluators, they were constantly changing. Moreover, access tothe foreign exchange market was typically not free and gave rise to parallel gray andblack markets, with exchange premiums that sometimes were expressed in multiplesrather than percentages.

Capital markets, too, were full of distortions. Rarely were banks free to competeamong themselves to attract deposits by paying depositors a competitive rate of interest.Instead, interest on demand deposits was often prohibited by law or at the very leastkept close to zero, while interest on time and savings deposits was also held far belowwhat would be its competitive level. In countries with significant inflation the interestrates received by depositors were almost invariably negative in real terms, sometimes tothe point of virtual confiscation.

Moreover, with or without inflation present, there were also huge implicit subsidies tocertain classes of borrowers. Sometimes the central bank itself gave subsidized credits,

51

52 Arnold Harberger

sometimes it was other public sector banks, (for example, development banks), andsometimes it was the commercial banks themselves (operating under legislativemandates or under directives from the government or the central bank).

Then there was the tax system. With few exceptions the system of corporation incometaxation was modeled after those of the industrial countries, and gave rise, in principle,to a situation in which the marginal product of capital was higher in the corporate than inthe noncorporate sector. But this broad tendency was often obscured by so-called taxincentives, which gave preferential treatments of various kinds to favored classes ofinvestment. In a number of countries, the accumulation of tax incentives over the yearscreated a mare's nest of "special treatments," causing huge distortions in the way capitalwas allocated among industries, sectors, regions, and specific uses. In the process, itvirtually wiped out the revenue-generating capacity of the corporate income tax.

Other mechanisms besides import protection, subsidized credit, and tax incentiveswere used to stimulate production or use of particular products, so one clearly hadproduct market distortions side by side with trade distortions and capital-marketdistortions. All these were quite well understood and reflected in the mainstreamliterature on project evaluation in the 1960s and early 1970s.

The project evaluation literature that appeared during this period was a naturaloutgrowth and, in a sense, a reflection of the policy environment just described. Withdistortions so rampant in the main markets of the economy, actual market prices wererather far from reflecting what standard competitive economics says they should reflect:the true economic cost of the good or service in question. This leads naturally to anattempt to quantify what these true costs really are. Thus the profession arrived at whatwere variously called the shadow prices, the economic costs, or the social opportunitycosts of foreign exchange, of capital, sometimes of main commodities like oil, and oflabor (about which more will come later).

To give an idea of how these social opportunity costs came into play, consider aforeign exchange market with three classes of imports (M1 with 100 percent tariff, M2with 50 percent tariff, and M3 with zero tariff) and three classes of exports (X1 with a 30percent subsidy, X2 with a 10 percent tax, and X3 with no special treatment). Theconsensus solution in the literature of the time was that the social opportunity cost offoreign exchange would be the relevant weighted average of the effective individualsocial opportunity costs applicable to each "source" of foreign exchange. Thus if it wasdeemed that an added dollar of demand for foreign exchange would be "sourced" 30percent at the expense of M1 , 20 percent at the expense of M2, 10 percent at the expenseof M3 and that the rest would come from newly generated exports-20 percent from AX1,and 10 percent each from AX2 and AX3-this would yield a social opportunity cost offoreign exchange of 14.5 pesos per dollar, as compared with a market exchange rate hereassumed to be 10 pesos per dollar. The calculation follows:

Project Evaluationfor the Next Decade 53

Weight (fi) DistortionSourcingfrom (percent) (percent) SOci fixSOCi

AM1 30 100 20 6.0

AM2 20 50 15 3.0

AM3 10 0 10 1.0

AX1 20 30 13 2.6

AX2 10 -10 9 0.9

AX3 10 0 10 1.0

Xj fiSOCi = SOC of foreign exchange = 14.5

Without going into detail about the conceptual foundation and the practicalestimation or assignment of the weights, there is a clearly huge difference between themeasured social opportunity cost of 14.5 pesos and the market exchange rate of 10 pesosper dollar.

Let us now shift gears and, using the same weighting structure, apply a set of traderestrictions that comes closer to the targets that countries are setting for themselves in the1990s.

Weight (fi) DistortionSourcingfrom (percent) (percent) SOCi fixSOCi

AM1 30 20 12.0 3.6

AM2 20 10 11.0 2.2

AM3 10 10 10.0 1.0

AX1 20 0 10.0 2.0

AX2 10 -10 0.9 0.9AX3 10 0 1.0 1.0

Xji fiSOCi = SOC of foreign exchange = 10.7

Here we get a weighted-average social opportunity cost of foreign exchange that isquite close to the market exchange rate of 10 pesos.

The main point of this example is that the adjustment needed to account for distortionsin the sourcing of foreign exchange is much smaller (in a typical developing country)today than it was twenty or thirty years ago. I believe these matters must still be treatedin a course aimed at the next decade, but the fraction of time devoted to them should bemuch smaller than before.

To buttress this point, consider the issue of expected changes in the real economic costof foreign exchange over time. Recall that a typical investment project will have anexpected life of twenty years or more, and that we must project the profile of costs and

54 Arnold Harberger

benefits of such a project over its lifetime.' There is little doubt in my mind that theexpected time path of market real exchange rates is worth at least as much attention asthe distortions that separate the social opportunity cost of foreign exchange from themarket exchange rate. Yet many courses in social project evaluation devote no attentionto this topic. Here is an indication of the volatility of the real exchange rate for a numberof countries.

Table 4-1. Real Exchange Rates, Selected Countries, 1970s, 1980s

RER indices (1985 = 100). 1970s 1980s

Country Highest Lowest Highest Lowest

111 United States 143 104 143 100

132 France 96 82 100 81

134 Germany 92 76 100 81

146 Switzerland 133 82 100 76

158 Japan 153 96 115 72

233 Colombia 118 90 154 81273 Mexico 122 90 159 86248 El Salvador 155 124 237 100299 Venezuela 113 84 184 73652 Ghana 235 57 235 18694 Nigeria 305 169 512 90636 Zaire 72 26 103 40

Basic Principles of Project Evaluation

1. The Profile of a Project (Ft = Bt - Ct)

a. Cash flow as a simple example and model.b. What to do when there are noncash benefits and costs.c. Dealing with taxes and subsidies when measuring natural economic benefits

and costs.d. Introduction to the least cost principle: Do not attribute a benefit that is

greater than the least alternative cost of achieving the same or similar benefit.

1. Perhaps the most common and egregious error in standard project evaluation over the yearshas been the failure to deal with prospective changes in relative prices. This error is reflected inreports that delineate in great detail the path of investment outlays over the gestation period andthen blithely project the "typical year" of operation, once the project gets going. There are surelynumerous cases in which the consequence of this error is mild, but they cannot be known inadvance without facing the problem of changing relative prices. The most critical items of changingrelative prices are in cases where the present price is out of line with the long-run expectedequilibrium prices.


2. Expressing the Profile in Real Termsa The basic processes of the economy are real processes.b. Economic analysis may help us predict inflation one or two years ahead, but

it is totally useless for ten or twenty years ahead. Factors like the strengthand responsibility of future governments are the critical determinants.

c. Economic analysis can help us project the real economy over a ten- ortwenty-year horizon-certainly more so than projecting the general pricelevel.

d. Projecting the economy "in real terms" includes projecting the course ofrelative prices. It simply makes no sense to project "at 1990 prices" keepingeach and every price at its 1990 level. The right way to project the economy isto project the course of individual prices, pj, relative to the general pricelevel, Pa. This price level should be quite general and give due weight tonontradables. Good candidates are the consumer price index (CPI) and thegross domestic product (GDP) deflator.

e. Examples of commodities whose prices are currently "out of line" with theirlong-run equilibrium levels-grains after the "Russian wheat deal" in 1974,sugar reaching 400 and 50¢ a pound in world markets (three occasions in the1970s and 1980s), copper falling below the marginal operating cost of mostmines (early 1980s). How to project a time path by which they will onceagain reach "normal" levels.

f Projecting the "normal" time path of relative prices. The trend of any givenrelative price reflects a battle between (hopefully) rising real wages on theone hand and technical advances specific to that commodity (or service) onthe other. Things like haircuts and taxi rides have little room for technicaladvance, hence they are observed to rise in real terms (so long as the labormarket yields rising real wages). Items like television sets and computershave enjoyed technical advances strong enough to more than offset risingreal wages, hence they have a downward trend in their respective relativeprices.

g. Projecting the time path of real wages. Real wages for a given class of labortend to reflect the overall average rate of technical advance in the economy,modified by supply and demand forces specific to the labor market inquestion. The premiums for different skills basically tend to reflect the returnon various kinds of human capital investment, but these in turn are onlyloosely connected to market interest rates and to the "capital market" assuch. Hence the skill premiums can change significantly over time.

h. Projecting the time path of the real exchange rate. The real exchange rate isbest thought of as the "real price of the real dollar." If the nominal exchangerate is E, then (E/pa) is the real price of the nominal dollar. If we can find orgenerate a world price index, ideally one that concentrates on tradable goods(call it px), we can get to the real exchange rate (E Px/Pa). We can track thisthrough time and analyze its movements to help us project its likely futurecourse.

56 Arnold Harberger

3. Finding a Real Interest Ratefor Discountinga. There are two important messages here: (1) the relevant interest rate for

discounting real flows must be a real discount rate, and (2) just as the realprices of other goods and services can and do vary over time, so too we mustallow for the possibility of a real discount rate varying over time.

b. It makes most sense to try to estimate the normal, real equilibrium rate ofreturn on capital for the economy in question and use that as the basis forderiving the relevant real rate of discount. The real discount rate is probablybest based on estimated patterns of sourcing in the capital market, but formost poor countries I would be prepared to simplify and use the real, gross-of-tax, measured rate of return to capital in the market-based portion of theeconomy. (More on this subject will come later.)

c. We must make allowances for cases where the current situation is one ofabnormal stringency (or glut) in the capital market. In such cases the realdiscount rate should start out higher (or lower) than the normal rate, andapproach the normal rate over time as the circumstances giving rise to theabnormal situation pass away or are surmounted.

d. Another class of cases that one might have to deal with for some countries isthat in which, because of deficient policies or external events, the observedreal rate of return on capital has been low for a considerable period of time. Ifthe policies have been dramatically reformed or the external causes of lowreturns have disappeared, there may be reason to set the forward-looking"expected normal real return on capital" at a level significantly higher thanthe recent observed level.

e. The best way to incorporate a changing rate of discount is to consider thediscount factor for any year t to be j=, (l + rj). That is, to discount Fl back toyear zero, we take F1 /(1+rl); to discount F2 we take F2/(1+rl)(l+r 2 ); for F3we take F3/(1+rl)(1+ r2)(l+r3) and so on. Here rl is the discount rate linkingyear zero to year one, r2 the one linking year one to year two and so on.

f It is very treacherous to work back from observed nominal interest rates infinancial markets, and to try to derive real discount rates for future flows onthis basis, for various reasons:(i) Financial interest rates can be tremendously volatile. There are countries

in which correcting such rates for the ongoing inflation yielded negativereal rates for significant periods. Yet in some of the same countries indifferent periods the same procedure produced real rates of 2, 3, even 4percent per month.

(ii) Financial interest rates do not incorporate the "standard" capital marketdistortions, such as property taxes and corporate income taxes. Nor is iteasy to "correct" them for this deficiency.

(iii) Financial interest rates are often subject to formal and informal controls,which are distortions in their own right, but not as easy to take intoaccount as tax rates.

g. For most developing countries, a measure of return to capital in the privatesector is the best starting point. In the numerator one has the total real returnaccruing, say, to private sector capital. In the denominator one has theprivate sector capital stock, built up from gross investment data using the


perpetual inventory method, with assumed depreciation rates. Ideally,separate calculations are done for buildings, machinery and equipment, andinventories. If the data are drawn from the national accounts, the numerator(return to capital) will likely include the returns accruing to land, while theperpetual inventory method (which accumulates gross real investment)yields an estimate only of reproducible capital. Thus one must eitherseparately estimate the return to land and take it out of the numerator, orseparately estimate the real value of land for each year and insert it into thedenominator. Another problem that arises using national accounts andsimilar (for example, business census) data is that the income ofunincorporated enterprises commingles the labor income of the owners (andtheir family members) with the return to capital of the firm. When this is thecase, one must resort to some system of dividing the global unincorporatedenterprise income into these two parts.

4. Types of Project Profilea. The project profile should in principle summarize all the relevant

information about the project. That is, it should capture all the relevantbenefits and costs.

b. This means that taking a different "point of view" usually means having adifferent profile. At least three points of view can be distinguished for astandard commercial project:(i) Owner's point of view. This follows a straight cash flow approach. Cash

coming in is a plus; cash going out is a minus. Receipt of a loan istherefore a plus; both interest and amortization of the loan are minuses.Depreciation of physical assets is not taken year by year but is reflectedin the difference between the cash outflow for asset acquisition and thecash inflow from disposing of the asset at the end of its useful life withinthe project.

(ii) Banker's point of view. The banker looks at a project before he decideswhether to lend money for it. As he analyzes the project, he does notknow how much of the financing will be debt or how much will beequity capital. Hence he looks at the full capital outlay as the investmentand the full return to capital (not saying which part goes as interest andwhich as return to equity) as the reward. The banker's point of view thusmerges or consolidates the accounts of those who provide debt capitaland those who provide equity. The banker's point of view is nonethelessa private one. It does not consider taxes generated by a project to bebenefits or make any other adjustments (for example, for externalities orsocial opportunity costs) that would be incorporated in an economic(social) analysis.

(iii) Economic (social) point of view. This point of view counts as benefits allthat the banker counts, but in addition adjusts for taxes, subsidies, andother types of external effects. The use of social opportunity costs (forexample, for foreign exchange) is simply a way of dealing with sucheffects for situations that occur repeatedly. Thus, the act of buyingforeign exchange causes displacement of "existing" imports and the

58 Arnold Harberger

expansion of exports. Tariff revenue is lost on imports that are displaced,as is the extra money spent as subsidized exports expand. Since theseeffects occur essentially whenever anybody enters the market to buyforeign exchange, we calculate their impact once and for all and reflectthat effect by simply using the social opportunity cost of foreignexchange as the cost to attach to any net purchase, or the benefit to attachto any net sale of foreign exchange by the project in any given period.

These uses of social opportunity cost are invariably connected withthe "sourcing" of the item in question. If the foreign exchange is spent ona tariffed item, the amount of the tariff should appropriately be countedas a benefit in the derivation of the project profit from the economicpoint of view. An example will probably help. A project needs imports of$1 million of equipment. It buys the dollars in the open market at 10pesos and in addition pays a 20 percent tariff. Total private cost: 12million pesos. This would be the cost from either the owner's or thebanker's point of view. However, from the economic (social) point ofview the cost would be 10.7 million pesos (assuming the socialopportunity cost of foreign exchanges is 10.7 pesos per dollar, as in theexample presented earlier). This can be represented as the private cost of12 million pesos, less the 2 million of import tariffs paid by the project (acost to its operators but a benefit to the treasury), plus 700,000 pesos ofdifference between the SOC and the market price of foreign exchange,(10.7-10.0) x $1 million.

The economic analysis of a project would treat in an analogousfashion the indirect taxes paid on a taxed input (for example, cement). Inprinciple, the negative externalities of air and water pollution stemmingfrom the project's emissions and effluents should be counted asadditional external costs in passing from the profile of a project from,say, the banker's point of view to the corresponding profile calculatedfrom the economic point of view.

Additional "social benefits or costs." Other considerations are sometimesbrought into the picture to evaluate government projects. There may be adesire to weigh the benefits and costs perceived by one group more heavilythan those perceived by others. Sometimes the meeting of specific "basicneeds" is given greater credit than that which is implicit in a market-priceanalysis. Sometimes the employment of the otherwise unemployed isdeemed to signal an extra benefit, not already counted in the calculus.

I know such considerations do in fact motivate policymakers in manyplaces. I know, too, that economic analysis alone will not tell us when andwhether to assign such a benefit or cost. Furthermore, a realistic assessmentof the considerations that actually motivate the authorities will reveal thatthey are not always high-minded and noble. They often amount to little morethan rewarding one's (political) friends and punishing one's enemies.Moreover, even when the motivation is pure, the quantification andcalculation of the benefit or cost to assess is almost invariably arbitrary.

My suggestion at this stage of the game is to be ready to quantify suchbenefits and costs but to label them explicitly as "project subsidies" and


"project taxes." Thus if I consider that 10 percent of the wages bill of aparticular project is a positive externality because of improved nutrition ofthose workers' children, I would count the full wages bill as a cost andintroduce two other entries. In the social evaluation I would add a benefitequal to 10 percent of the wages bill and call it a child nutrition benefit. In thefinancial analysis of the project I would add a corresponding item labeled"project subsidy to child nutrition."

This procedure keeps the concepts straight, gets implicit subsidies moreout into the open, and immediately reveals when a project ends up passingthe critical test only because of these ascribed social benefits, or ends upfailing the test because of similarly ascribed social costs.

5. Working with the Project Profile (internal rate of return, net present value, and soforth)

a. Calculation of the internal rate of return (IRR): The project profile can beevaluated at different discount rates. Each rate will yield its own net presentvalue (NPV) calculated to a given point in time. As long as the project has anormal profile (with periods of net outlay first, followed by periods of netbenefits), there will be only one rate that generates an NPV of zero. That rateis the project's internal rate of return. Its great merit is that it is an attribute ofthe profile as such; to calculate it one does not have to know the relevantopportunity cost of capital (discount rate).

b. Internal rate of return is not a criterion.Simple counterexample:Project A: You lend me $1; I pay you $2 one year hence.Project B: You lend me $1,000; I pay you $1,250 one year hence.In the U.S. capital market context, B is clearly preferable to A, yet A has anIRR of 100 percent, B of only 25 percent.Suppose the opportunity cost is 10 percent. Then project A, evaluated as ofthe endpoint, has NPV = 90¢. But project B, evaluated at the same point intime, has NPV = $150. This is why we prefer B. If the relevant discount ratewere 30 percent, B would not be interesting. Its NPV would be negative.In this example A and B must be seen as strict alternatives. Otherwise, wedon't have to choose between them; we can do both.But it is not artificial for A and B to be strict alternatives, and for one to costmuch more than the other-for example, a high dam or a low dam at a givensite; a gravel road or a concrete road connecting two towns. One shouldthink of the IRR as a useful descriptive statistic, nothing more than that. It isdefinitely not a criterion. (Sometimes, indeed, there are multiple IRR, but thisis not common in real-world projects. This never happens with profiles of"normal" shape.)

c. Net present value is the relevant criterion as long as the discount rate reflectstrue opportunity cost.If you can earn 10 percent on your money in a bank, you could do better byinvesting in a project whose NPV, calculated using 10 percent as the discountrate, is greater than zero. NPV is, in fact, a measure of the profit you makecompared with keeping your money in the bank, eaming 10 percent.

60 Arnold Harberger

d. If the relevant opportunity cost of funds is expected to vary through time,NPV is still the correct measure. But now NPV must be calculated at adiscount rate that varies through time, as discussed above under (3e).

e. The benefit/cost ratio is even less a criterion than IRR.

Project APresent Value, All Benefits 1,000Present Value, Capital Costs (C1) 300Present Value, Current Costs (C2) 500

Project BPresent Value, All Benefits 1,000Present Value, Capital Costs (C 1) 500Present Value, Current Costs (C2) 250

If benefit-cost ratio is defined as PVB/(PVC1 + PVC?)we find project B > project A.

If benefit-cost ratio is defined as (PVB - PVC?)/PVC1we find project A > project B.

Both these procedures are used in actual cost-benefit analyses.The NPV criterion gets the same answer when it compares total benefits

with total costs as when it compares net current benefits with capital costs,that is:

PVB - (PVC1 + PVC2) = (PVB - PVC2 ) - PVC1

The difference is fundamental. Net present value is fundamentally anadditive concept. It works in both cases because of the commutative law ofaddition.

The benefit/cost ratio in a sense founders because it cannot deal with afundamentally additive concept.

6. Basic Exercise 1: The Timing of a ProjectOne of the most basic cases of one project being an alternative to the other occurswhen it is essentially the same project but is executed at different times.

To deal with the problem of timing, we must realize that to compare projects withdifferent timing we must discount (or accumulate) their net present value to the sameyear. It can be year 0, 1, 2, or k but it must be the same year for all projects beingcompared.

One can think of starting the project at many different possible times. For eachpossible starting date we select the "best" design for the project, done at thatparticular date. We get a profile for that project (in principle) and for the "best" designfor each alternative starting date. Then we compare the net present values of theseprojects, taking present value to a common date. The one with maximum NPV is theone with the optimal starting date.

The above procedure would be tedious and require lots of information about whatproject benefits and costs would look like for different starting dates. Hence it is not


likely to be done with the proper care. We should seek shortcuts for the analysis oftiming.

But one thing remains clear, even from the above quick sketch. If we have a curveshowing how NPVJ (net present value of the project started at time j, with NPV

measured to time zero), we want to choose j to be the time for which that function isat its maximum.

Now, the typical project probably starts out as "not yet ripe" for the picking. Muchof the time, the project will have a negative NPV if done right away. Common sense,as well as our analysis, counsels us to wait and see if NPV turns positive at some laterdate. So we, the technical people, tell the politicians and administrators and theinterested parties that they should wait. We probably even show them that NPV isnegative and convince them they must wait for that reason. So they wait, perhapsseveral years. Finally a year comes when the NPV finally turns positive. Now thepressure gets really severe. Unless we believe that the maximum NPV for the projectis at or near zero, we should counsel everybody to keep on waiting. Even thoughNPVJ is now positive, it is probably not at its maximum.

I feel certain that, as a general rule, projects are "timed" too early, precisely for thereason implicit in the above scenario: People want to do the project as soon as it willpay off-maybe even sooner. (The great bane of project evaluation is interestedparties who enjoy most of the project's benefits while paying only a fraction of thecost. From their narrow point of view benefits are always greater than cost-that is,the part of cost that they pay. They cannot imagine that the project is not good for thewhole society, just as it is so clearly and obviously good for them. Think of thefarmers in the region of an irrigation project. They get nearly all the benefits and maynot pay even 2 percent of the cost borne by the public treasury.)

It is critical to instill in the minds of participants in project evaluation courses thatthere is a very strong presumption against building projects at the first point in timewhen their NPVj turns positive.

7. Basic Exercise 2: A Special Case of TimingThere is a neat special case when (a) project benefits, Bt, are an increasing function oftime and (b) when the benefit of a given year, t, is the same regardless of when (priorto t) the project began. This pair of assumptions is pretty accurate for road projects.Since traffic seems "everywhere and always" to increase with time, and since thebenefits of a road project are closely and positively associated with the traffic on theroad, it is overwhelmingly likely that the time path of benefits will be upward. Also,with a bit less certainty, the traffic carried by a given road today would not be muchdifferent if the road had been built (or upgraded to its present specifications) five orten years earlier.

In this special case, and with capital costs of the road or road improvement beingconstant through time, the timing rule is to build when the first year's benefit firstexceeds the forgone interest.

The idea is that each year of waiting allows society to enjoy a yield equal to theopportunity cost of capital (r) times K*, the amount of funds involved. Therefore, aslong as Bt+1 is less than rK, it pays to wait. As soon as Bt+l is greater than rK, it paysto build.

62 Arnold Harberger

Note that this theorem deals with the first-year benefits only. Obviously if the first-year benefits are bigger than rK, and if benefits grow through time, every year'sbenefits will be bigger than rK. It is easy to show that this rule gives us a maximum ofNPVJ ; hence it does not fall into the trap pointed out at the end of point (6).

We can modify this simplest special case to take into account capital costs thatchange through time, and to be less rigid about Bt always increasing through time.When capital costs change through time, that is, Kt+l # Kt, it is reasonable to supposethat the movement is unidirectional-either up or down-over a period of years. Inthis case the rule becomes "Build when Bt+l first exceeds rKt - (Kt+l - Kt )" instead of

"Build when Bt+l first exceeds rK." If capital costs are growing, this modified ruleleads to building the project earlier. If capital costs are falling, it leads to furtherpostponement (in order to take better advantage of the falling costs).

The important aspect about the annual benefit stream Bt is that it should be risingin the neighborhood of the optimal date. If later on it starts to fall, this does notmodify the optimal date at all.

What it does do is give rise to the possibility that even constructing at the optimaldate given by the relevant rule, NPVJ is still negative, even at that optimal date. Thusif we think it plausible that Bt although not rising, will tum down later on, we followthe formula to get the optimum time of construction in the regular way. Then,however, we must perform an added check of calculating the net present value of theentire project to make sure it is positive.

8. A Digression in Successor Projects and Opportunity CostWhat happens when you compare altemative projects having different economiclives? Is it "unfair" to make the short-lived project compete with a longer-lived one?

More generally, what about the terminal dates of any and all projects? Might it notbe appropriate to consider not just project A but a sequence including A and itssuccessors, A', A", and so forth?

There is one important and very basic message here. Normally we expect successorprojects, and any other projects whose identity is unknown, to have a marginalproductivity of capital that is equal to the relevant opportunity cost of capital.

This message derives almost from the definition of opportunity cost. If theopportunity cost of capital is 10 percent, we should do all independent projects thatyield more than 10 percent. The only time we should leave undone a project withyield higher than 10 percent should be when that project has been beaten out by abetter altemative. In that case it is obviously not an independent project.

If, on the other hand, we have vast numbers of projects with yields higher than 10percent that we cannot do for lack of funds, this practically tells us that the relevantopportunity cost is higher than 10 percent. It may be that we have simplymisestimated the opportunity cost and that it should be raised straightforwardly to 12or 15 percent.

In some situations, there may be a transitory stringency of funds because of specialfiscal problems, bad crops, or even the simple fact that a large project is making hugebudgetary demands and so little money is left over for other projects, even good ones.

In such cases it is appropriate for the relevant discount rate to be raised above 10percent for the present period and for as long into the future as the special stringency


of funds will prevail. If this adjustment is appropriately done, it should be possible toapprove all independent projects that meet the NPV > 0 test at the revised set ofopportunity cost rates.

Returning now to successor projects, the normal presumption is that these as yetunknown projects, taking place 10 or 20 years hence, will be yielding benefits at thestandard opportunity cost rate. If this is so, NPV will be approximately zero and wecan duly neglect them. Even more so when we realize that to incorporate A'(beginning, say, in 10 years' time) and A" (beginning 20 years from now), we wouldappropriately have to discount their NPVs back to the same time we are using as thereference point for A.

The preceding discussion gives a clue as to when we should think seriously aboutsuccessor projects. This occurs when the successor project promises not to be amarginal project but a supramarginal one. The natural scenario for this is a roadproject. When the old road reaches the end of its economic life, its demise can give riseto an opportunity for a new successor project that is supramarginal in the sense ofhaving an IRR well above the opportunity cost rate and consequently, a significantlypositive NPV, measured using the opportunity cost rate for discounting.

When this happens, it is appropriate to include the successor project or projectsalong with the specific new one being started. That is, we should treat (A + A') or(A + A' + A") as the profile to be analyzed, instead of A above.

9. Basic Exercise 3: Comparing Projects of Different LivesSuppose A (an asphalt road) has an economic life of 15 years, while C (a concreteroad) lasts 30 years. This is a simple case to deal with, assuming that A's successorproject also has a life of 15 years, for then one compares NPVo of C with the sum ofNPVo (A) and NPVo (A'). Each alternative being compared runs out in year 30, so wehave no need to introduce any alternatives after that. We are open at year 30 to chooseC' (the successor to C) or A" (the successor to A + A'), regardless of which option wetook up to year 30.

The problem comes when A has a life of 15 years while C lasts 25 years. Now, inorder to make the sequences come out exactly even, we would have to go 75 years inthe future. And it could get a lot worse-if A's life were 17 years and C's 29 years, wewould have to wait 493 years before the two sequences ended at exactly the sametime.

It is obviously nonsense even to think in such terms, but if we are to avoid thinkingthat way, we need some device to help us. It is convenient here to think in terms oftrying to be as fair as possible to each of the two alternatives.

If we simply compare NPVo (A) with NPVO (C), we are implicitly assigning a netpresent value of zero to A's successor project. This is being unfair to A and biasing thechoice in favor of C if we assume both A and C have significant positive NPVs.

Alternatively, if we compare C with A + A', making no adjustment, we are biasingthe result in favor of A + A' by implicitly assigning an NPV of zero to project C'.

The trick is to create an artificial project, Ai, which has NPV • 0 and terminates atthe same time as C. Suppose C has NPVO = 250 while A has NPVO = 200. Suppose, too,that Project A' runs from year 16 to 30 and has NPV1S = 200. Let the discount rate besuch that NPVo (A') = 100. Thus, in the unfair comparison favoring C, C would winout because 250 > 200. In the unfair comparison favoring A (taking the full NPV of

64 Arnold Harberger

A + A' and comparing it with that of C) the sequence A + A' would win because(200 + 100) > 250.

To mediate this impasse, we suggest calculating the internal rate of return of A'and dividing its benefit profile into BF (= benefit stream up to year 30) plus BG(benefit stream after year 30). We know that the present value of costs equals thepresent value of benefits when p, the IRR of A', is used for discounting. Hence thepresent value of costs of A' is equal to PVBF ( p ) + PVBG ( p ) we can therefore assign asum equal to PVBF ( P) as the present value of the cost of producing the benefit streamBF. Together this cost and benefit stream make up the artificial project A' F. It has apositive net present value when r (the opportunity cost discount rate) is used, but sotoo does A'G, which is the part of the profile of A' we have left out. We have not beenfundamentally unfair either to (A + A') or to B in adopting this device for splitting thecosts and benefits of A'.

The above is not a profound theorem, but rather a "trick" to help surmount anotherwise troublesome problem. As a child, I and surely millions of children aroundthe world, was fascinated by the box in which Mother's Oats cereal came. The box hada picture of a mother holding a box with a picture of a mother, who in turn washolding a box with a picture of a mother and so on. I remember getting out amagnifying glass to see how many mothers I could discern on that label. A trivial andfruitless enterprise, a waste of time, perhaps also misplaced childhood imaginings!Whatever the case, it is similar to the project evaluation problem when twosupramarginal alternatives of unequal lives have to be compared. Our trick isdirected at helping analysts surmount their fascination and get on with the seriouslabors they have before them.

10. Basic Exercise 4: The Problem of ScaleThe analysis up to this point carries, both explicitly and implicitly, the message thatthe guiding principle of project evaluation is the maximization of net present value,using the relevant opportunity cost of capital as the rate of discount. Most scenariosand examples in the end boil down to being applications of this general principle. Thepoint of the exercises is not to change the principle but to show people how to applyit, to give them an intuition concerning what sorts of problems (conceptual andpractical) may arise, and to lead them to new ways of thinking about differentapplications.

The problem of scale is a case in point. Here many alternatives are possible, but(practically by definition) they are ordered in terms of their capital cost, fromsmallscale up to largescale.

This characteristic gives rise to the possibility of comparing increments of cost toincrements of benefit as one moves from the design for a smaller-scale operation tothe design for a larger-scale one. Indeed, one can think of a profile built up fromincrements, with the incremental cost coming first, followed by years of operation inwhich incremental benefits are produced.

Working with such a profile of increments, we can derive its net present value,using any desired interest rate, and obviously also its internal rate of return. Usingthese concepts, we can find some interesting and useful results with respect to scale.

a. If the relevant discount rate is 10 percent, there is obviously no point inchoosing a scale that has IRR = 10 percent, for at such an IRR, NPV will be zero.


If we maximize NPV using a 10 percent interest rate, the resulting scale ofproject should have an IRR > 10 percent.

b. But while IRR > 10 percent at the optimal scale, MIRR (the marginal orincremental IRR) will tend to be equal to 10 percent. Think of proceeding fromthe smallest scale by increments. If the first step has incremental benefits > cost,using 10 percent, then we should take that step. But also its MIRR will be > 10percent. We should continue to do this through successive steps as long as theincremental NPV is positive. All such successive steps will have MIRR > 10percent. Where does this process stop? Naturally, where the next step will havea negative incremental NPV, and hence an MIRR < 10 percent. Basic conclusion:Optimizing the scale of a project typically means expanding scale until themarginal internal rate of return is equal to the opportunity cost of capital. Thismaximizes the NPV of the whole project in using opportunity cost as thediscount rate.

c This juxtaposition of NPV, IRR, and MIRR is a unique attribute of the problem ofscale. If 10 percent is the discount rate, we go to the scale for which MIRR < 10percent. That is also the scale for which NPV(10 percent) is at its maximum. Solong as NPV(10 percent) is positive, the IRR at that scale will be greater than 10percent.

If 12 percent is now the discount rate, we go to the scale for which MIRR is 12percent. That is also the scale for which NPV(12 percent) is at its maximum.Again, typically IRR will be greater than 12 percent at this point.

d. If the relations involved can be approximated by continuous curves, then foreach discount rate below a critical level there will be two scales of the project forwhich that discount rate is the IRR. This is necessary because if the maximumNPV(12 percent) is at 812, then there must be some S12 scale less than S12,where NPV is zero (and rising with scale) and 812 another larger than 812,where NPV is zero (and falling with scale). There will, however, be just onediscount rate (call it h) for which NPV = 0 at 3(h). That is the highest discountrate under which this project could survive as acceptable. But its scale S (h) isnot the right scale at which to build the project. With each reduction of thediscount rate below h, the optimal scale changes (presumably increases) assuccessive increments to scale that did not pass the test at higher discount rateswill come to be acceptable as the discount rate is lowered.

11. Basic Exercise 5: Interrelated ProjectsAs in economics in general, we distinguish in project evaluation the relationships ofsubstitution, complementarity, and independence. They are defined for the sides ofbenefits and costs as follows. Projects I and II are complementary on the benefit sideif

PVB(+II) > PVBI + PVBII.This says that the benefits of doing the two projects "together" are greater than thesum of the separate benefits that would accrue from each one done separately.Similarly, projects I and II are substitutes on the benefit side if:

PVB(I+n) < PVBI + PVBII.In this case "something is lost" by doing the two projects "together."

66 Arnold Harberger

Finally, projects I and II are independent on the benefit side ifPVB(1+I) = PVBI + PVB1 I.

Here it simply does not matter, as far as benefits are concerned, whether the projectsare done "together" or separately.

I put quotation marks around the word "together" because it does not mean theyhave to be done simultaneously. They may not even be located near each other. Theidea of "together" means that both projects are simultaneously present (on the scene)when (most of) their benefits are being generated.

Corresponding to the three definitions above, we have:Complements on the Cost Side

PVC(Q+II) < PVCI + PVC 1i

Substitutes on the Cost SidePVC(I+II) > PVC1 + PVCj1

Independent Projects on the Cost SidePVC(I+l) = PVCI + PVC1 I.

Note that the inequalities on the cost side go in the opposite direction from thoseon the benefit side. This reflects that complementarity means there are advantages indoing the two together-this happens when one gets greater benefits by doing themtogether, or lower cost. Similarly, substitute projects reflect either smaller benefits fromdoing them together, r, or greater costs. Recalling that the ultimate objective is netpresent value is another way to see why the definitional inequalities are in theopposite direction, for NPV = PVB - PVC.

12. Examples of Interrelated ProjectsSubstitutes on the Benefit Side

I = a power damII = an irrigation dam(1+11) = a combined power-irrigation dam

Many people think the two uses of the dam are complementary. This is true on thecost side but not on the benefit side. The simplest way to see this is to recognize that ifthe dam is only for power, the electricity agency will be able to release water in apattern that is totally optimal from the standpoint of energy operations. Likewise, ifthe dam is only for irrigation, the agricultural ministry will release water in thepattern that is optimal for agriculture. When we try to use the dam for both purposes,it would be pure serendipity if both "managers" would want exactly the same timepattern of releases.

In reality this will never occur. The electricity authority will want to use the waterat times of daily and seasonal peak demand. The agricultural ministry will want touse the water in the particular weeks when the crops need it most. If the project is ajoint power-irrigation project, therefore, one or both of the "managers" must take lessthan he would ideally like. This is what guarantees

PVB(1+l1) < PVB(J) + PVB(II).The same sort of story applies with all kinds of multipurpose projects. Both the

power use and the irrigation use think of the water as something to be used. In thecourse of a season the dam might go from full to empty, or nearly so. These uses aretherefore in conflict with the goals of a flood control objective, which would want tokeep the dam empty most of the time to cope with an unexpected flood, or with a

Project Evaluation for the Next Decade 67

recreational objective, where the goal would typically be to keep the water level moreor less constant to make good use of beaches, boat docks, and moorings.

In general, one can expect that most multipurpose projects will be substitutes onthe benefit side.Complements on the Benefit SideThese tend to be projects such as:

I = an airport projectII = a highway project improving access to the airportI+II = simply having I and II "together"

These are projects whose benefits are larger with larger traffic volume. Thepresence of one of the projects adds to the traffic flow that the other can expect, andhence adds to its benefits. In most pairs of projects of this type, the presence of theone adds to the benefits of the other.

One famous but somewhat different example is the textbook case of an orchardand an apiary. Here the flowers of the orchard feed the bees, while the bees' activitypollinates the orchard. In this case the presence of each project augments theproductivity of the other.Complements on the Cost SideObvious cases are:

I =adamII = a highway project that can utilize the dam for crossing the

river and thus obviate the need to build a bridgeI = a logging or mining projectIf = a port project that, when done together with I, reduces

the transport costs from 1.Substitutes on the Cost SideHere the most obvious way one project can make another more costly is byaugmenting the transport costs of the second project. Any projects that involve largeland assembly with little or no traffic "through" the assembled land have thischaracteristic. Thus, the large lakes created by dam projects render impossible (ormuch more costly) any highway project that would have run where the lake now is.

They also render more costly other types of projects whose natural lines of supplyof inputs or shipping of products would have taken a direct route "through" what isnow the lake.

Large metropolitan airports are like the lakes created by dams in that they toorepresent new obstacles that traffic now has to by-pass usually in more expensiveway.

13. Basic Exercise 6: When to Replace an Old Asset with a New OneThis is a problem that every family and every business, from the smallest to thelargest, has to face on a rather regular basis. The full generality of the problem can beseen by considering a tiny enterprise that operates with just one truck.

One can assess the replacement option by calculating the NPV of the trade. This isNPVT = PVBN - PVBO + SVO - PN

Another important challenge is that of taking properly into account the probabilitydistribution of market prices and other important variables. The probabilitydistributions of the relative prices of most raw materials and raw foodstuffs are highly

68 Arnold Harberger

skewed, so their means are well above their most commonly observed values. Yetmost project evaluations simply use the commonly observed values as the "expectedvalue."

There are also probabilistic interactions among sets of prices' (for example, thegrains), and among prices and quantities in different markets. Analysts can take theseinto account using Monte Carlo techniques, but their use has been generallyrudimentary.

The above only sketches some of the challenges that still are present even if weassume society is neutral with respect to risk. Certainly these challenges are adequateto keep us busy for at least a decade or two!

But if that were the case for alternatives as close to each other as 3.1Y and 3.3Y, itfollows that society would have little interest in, say, doubling its income-somethingwe know is far from the truth.

The prudent conclusion to draw from the above sketch and from similar reasoningcarried out at greater length is that it is not at all implausible to assume that society'smarginal utility of wealth is approximately constant over the probabilistic range ofoutcomes associated with even a very large project.

14. Dealing with "Risk" in the Next DecadeSome may believe that the assumption of risk neutrality simply assumes away theproblem. It gives a result that seems to make the analyst's life easy, and thus looks likea lazy man's way out. Such a notion is far from the truth. The fact is that the art andscience of project evaluation are far from extracting the full meat out of the simpleassumption of risk neutrality.

As indicated above, risk neutrality implies looking only at the expected value ofthe outcome for each project. The bad news is that this is something that few if anyproject evaluations have so far succeeded in doing. Mostly, people work with vaguenotions of "most likely values" of the prices, quantities, incomes, wage rates, interestrates, and so on that enter into project evaluations. These may correspond to themedian or the mode of a probability distribution of outcomes, but not the expectedvalue. For example, I know of no evaluation of a dam that incorporated the hugepotential cost of the dam's eventually breaking. Admittedly, such a cost would enterthe expected NPV with a low probability, but the product of cost x probability mightstill be a quite significant evaluation. We do not even know how to define society'sportfolio. We wouldn't know how to begin measuring what a project like a dam or aroad did to the total variance of outcomes of society's portfolio. We have not theslightest idea how "society" judges the riskiness of the portfolio that we have not beenable even to define.

The good news is that we can go a long way without ever having to confront theabove conundrums. The key is to assume that society is neutral with respect to risk.So long as we can work with that assumption, we need only ask about each project:What is the expected value of the net benefits that it will generate?

It is rather easy to build a case that society should be treated as a neutral withrespect to risk, as far as social project evaluation is concerned. The main concept in theanalysis of risk preference and aversion is a function describing the utility attaching toeach level of wealth. A function in which the marginal utility of wealth declines aswealth increases displays risk aversion, and with increasing marginal utility of wealth


shows risk preference. A function that is a straight line over a certain range shows riskneutrality over that range.

Now try to draw an analogy to the above for society. Society's wealth is probably 2or 3 or 4 times national income in most countries. Hardly any project is larger than 10percent of national income. A large, risky project might have the outcome thatsociety's wealth is 3.1 times Y with 50 percent probability and 3.3 times Y with theother 50 percent probability. The social valuation of that package is not likely to differmuch from the social valuation of a wealth of 3.2 times Y for sure. In order to get theresult that society really values the risky option much less than the sure option, youwould need to have a sharply declining marginal utility of wealth.

15. Some Reflections on Allowingfor Risk in Project AnalysisThere are two quite different definitions of risk in the literature of finance andeconomics. One of these, default risk, is what we see in risk premiums on low gradebonds; its function is simply to allow for the probability that the expected paymentswill simply not be made. The second definition of risk is portfolio risk; it concerns thevariability of outcomes to which a given portfolio is subject.

When economists talk about risk aversion or risk preference or risk neutrality, theyare speaking of the attitudes of individuals toward the risk they see (or bear) on theiractual or potential portfolios. Most of the literature assumes that most marketparticipants are risk averse. The market itself seems to reflect risk aversion in the wayit prices many, though not all, assets.

It is not easy today to infer how risk is being evaluated in the market because it isnot true that risk is inherent in a particular asset. That was the old-fashioned and easyway of viewing risk. An asset's riskiness was simply measured by its coefficient ofvariation (standard deviation + mean) of income, appropriately measured. High-variance assets were risky, and low-variance ones were nonrisky.

The modern approach now looks at the entire portfolio-if a portfolio consists ofassets that are cyclically sensitive, one does not try to reduce risk by adding a newasset with low variance. Rather, one looks to find a new asset whose performance iscountercyclical. If such an asset can be found, high variance in it might well be a verypositive attribute, for it can then more powerfully offset the cyclical riskiness of theexisting portfolio.

The modem approach to risk was a big help to financial managers and advisers,but it is of no help to those who engage in social project

NPVK = PVBN/O PN > °

These inequalities tell a very clear story. If we are to keep the two assets, each mustjustify itself as the marginal asset. If one of them cannot, there is a better alternativethan having both assets simultaneously on hand.

16. The Principle of Separable ComponentsThe preceding exercise leads into one of the most important propositions in appliedcapital theory-the principle of separable components. This principle simply extendsthe lesson of the preceding exercise to any number of separable components. It statesthat for the best package of components, each and every separable component must

70 Arnold Harberger

justify its presence in the package, as if it were the marginal component. If it cannotjustify its presence in this way, the NPV of the package will be increased by simply notincluding that particular component.

This principle goes to the core of project design. Should a dam be 90 meters high or100 meters? The last 10 meters are a separable component (in the design stage) andmust justify their incremental cost with a corresponding incremental benefit. Shouldan irrigation project serve 2,000 hectares or 2,500? The last 500 are a separablecomponent and should justify their incremental cost with at least a matchingincremental benefit. Should the canals of the irrigation project be lined? Should ahighway be 6 meters wide, or 7? Should another highway have four lanes or six?Should a port have 2,000 meters of mooring space or only 1,500?

There is simply no end to the questions, generally aimed at the design stage of aproject, whose answers entail an application of the principle of separable components.

Where PVBN = present value of benefits of new assetPVBO = present value of benefits of old assetSVO = sale value of old assetP = price of new asset,

the calculation is straightforward to the point of being tedious.But this way of looking at the problem is not the only way, nor typically even the

most appropriate way. Typically, we must also consider the option of keeping the oldasset while buying the new. The present value of this option can be described by

NPVK = PVBO+N - PVBO - PN.Here PVBO+N = present value of benefits of "old plus new" assets together. If the

option of having both assets together is the best option, then NPVK must be greaterthan NPV T and must itself be positive.

Calculating out the differences, we findNPVK - NPVT = PVBO+N - PVBN- SVo > 0NPVK = PVBO+N- PVBO - PN > 0°

It is appropriate to definePVBO/N = PVBO+N - PVBN

and PVBN/O = PVBO+N - PVBOPVBO/N is the present value of the benefits from the old asset, given the new. It is

calculated by assuming the new asset will surely be there, and asking how much theold asset adds in the presence of the new.

PVBN/o is the present value of the benefits from the new asset, given the old. In itscalculation we treat the old asset as being there, and ask how much the new one willadd to the presence of the old.

Armed with these concepts we find that:NPVK - NPVT = PVBO/N - SVo> 0

Notes on Some Issues in Social Project Evaluation

Arnold Harberger

In "Project Evaluation for the Next Decade," I put forward the idea that we, in ourapplications of social project evaluation, are well advised to treat society as neutral withrespect to risk. I base this recommendation both on things we are pretty close to"knowing"-which tell us that neutrality is probably a rather good approximation-andon other things that I wonder if we can ever really "know" but that we would have toknow in order to apply an assumption, say, of risk aversion.

First, let us turn to what we more or less know. The economic analysis of an agent oran entity facing risky alternatives can be framed in terms of that agent's or entity's utility-of-wealth function, which expresses utility as a function of wealth. Where this function isa straight line, risk neutrality prevails; where it is concave upward, we have riskpreference; where it is concave downward, we have risk aversion. In what follows, I willassume that most people tend to be risk averse (demonstrated by their willingness to buyinsurance of all kinds), although one must realize that risk aversion is far from universal(namely, Monte Carlo, Las Vegas, Atlantic City, and the hundreds of national and statelotteries that do vast amounts of business every day, not to mention highly speculativestocks).

TIhe second step in the analysis of risk is to define the alternatives. Consider anindividual homeowner whose wealth is equal to twice his annual income (Y), all of itinvested in his house (H). Suppose that he can insure his house against destruction bypaying a premium equal to .03Y. We can then describe his options as being the following.If he does not buy insurance, he has H plus Y with probability .99, and zero plus Y withprobability .01. If he does buy insurance he has H plus .97Y with certainty. Note that inthis case the insurance is not a "fair" gamble-he pays a premium of .03Y to insureagainst a loss whose expected value is .02Y [- probability of loss (.01) x amount of loss(2Y)]. Thus the homeowner will buy insurance only if, faced with this choice, he is riskaverse.

Most homeowners do in fact insure their homes against catastrophic loss. Let us gothrough "their" calculation. If they pay the premium, they lose MUw, the marginal utilityof wealth, times .03 Y. If they do not pay the premium, their expected utility loss is theaverage utility of wealth (U/2Y) times the probable loss (.02 Y). Thus if one's ratio of themarginal utility of wealth to its average utility is less than 2/3, the solution is to buyinsurance; if one's ratio is greater than 2/3, the solution is to self-insure.

71

72 Arnold Harberger

Now consider the case of a TV set (T) that cost .04Y when new. The homeowner whohas bought the house insurance is now offered insurance against theft or calamity at apremium equal to 15 percent of the value of the set. Assume, too, that the actualprobability of loss is 10 percent, reflecting the same relation of premium to expected lossas we had in the house insurance case. What are the choices he faces? If he does not buyTV set insurance, he has H plus T plus .97Y with probability .9, together with H plus .97Ywith probability .1. If he does buy insurance, he has H plus T plus .97Y - .006Y withcertainty.

The utility calculus in this case runs as follows. If he buys insurance, his expectedutility is U(H + T + .97Y - .006Y). If he does not buy insurance, his expected utility is.9LU(H + T + .97Y) plus .1 U(H + .97Y). Here we have to compare the marginal utility ofwealth times .006Y with .1 times the difference between the utility of wealth "with" theTV set and the utility of wealth "without" the TV set. Call this difference the"incremental utility of wealth" times the value of the TV set. The "incremental utility ofwealth" is the slope of a chord connecting two points on the utility function-one ofthem with and one of them without the TV set. The question is whether the marginalutility of wealth is greater or less than 2/3 of the "incremental utility of wealth," wherethe incremental utility of wealth is measured over a span equal to approximately .04 Y, or.02W.

My point is that few households would choose to insure against the loss of a TV set,given the choice as stated. The reason is that there is little difference, over so small arange, between the "marginal" and the "incremental" utilities of wealth. Put anotherway, if we assume that these households have a constant marginal utility of wealth, wewon't make much of a mistake with reference to the TV insurance problem, but weprobably would make a significant mistake with reference to the house insuranceproblem.

The next step is to assert that for every case of social project evaluation that I can thinkof, the problem of risk analysis lies much closer to the TV insurance case than to thehouse insurance case. As I write this, news broadcasts are still full of the aftermath ofHurricane Andrew, with losses estimated at $20 or $30 billion. How does this comparewith "society's wealth" in the United States? With gross domestic product (GDP) inexcess of $5 trillion, and using a ratio of 3 or 4 between national wealth and nationaloutput, we have national wealth in the range of $15 to $20 trillion. Hurricane Andrew'slosses are equal to somewhere between .001 and .002 of total national wealth. To theextent they are covered by the taxpayers, the cost entails far less than 1 percent of thewealth of each of us. How can we worry about curvature in society's utility function forwealth over such a small range? Any curve can be approximated as a series of linearsegments. If we take a hypothetical utility of wealth curve for society and break it upinto, say, 20 linear segments, we will get a close approximation to the total curve. Nowwe come to my final point: my assertion that the risk issues surrounding each and everyproject I know of, and each and every relevant package of projects I know of, do not carryus off the last-the 20th-such linear segment.

Therefore, we are on quite solid ground in saying that a linear approximation tosociety's "utility-of-wealth" function is really a close approximation for the sorts ofvariations involved in the risk analysis of individual projects or relevant packages ofprojects. Accordingly, the assumption that society is neutral with respect to risk is valid,to an equally close approximation.

Some Issues in Social Project Evaluation 73

Portfolio Risk and Covariance

The problem of defining "society's portfolio" seems to be almost insurmountable. It isreasonably hard work to come up with estimates of society's overall real wealth. Themethods we use for this do not distinguish among the narrow categories of wealthrequired to talk seriously about portfolio risk and the covariance of alternative outcomesof a particular project with the corresponding values (movements) of society's portfolio.

The most widely used methodology for establishing society's wealth consists ofbuilding up a time series of capital stock, using the perpetual inventory method. This isbased on national accounts data on investment, broken down into categories likeresidential construction, nonresidential construction, machinery and equipment, andinventories. Rarely are these categories broken down by individual industry or by region;yet this is the type of breakdown one would need to make a good guess as to whether aparticular new project has a positive, zero, or negative covariance with "society'sportfolio."

My reaction to this difficulty is a strong "Why chase this particular rainbow?" Whyshould we spend vast amounts of professional time and resources on a quest that has solittle promise? Especially so in light of the fact that the main point of the previous sectionremains valid-that is, the assumption of risk neutrality yields a close approximation tothe truth in any case.

For readers unfamiliar with how covariance enters into the analysis of portfolio risk,table 5-1 presents a very rudimentary example. It shows five alternative "states ofnature"-Depression, Recession, Normal, Prosperity, and Boom. Row 1 shows theprobability associated with each state, while row 2 gives the value that the old portfolio isexpected to have under each of these states. Now the investor is looking at two assetsthat he might add to the old portfolio. I assume that each asset (A and B) costs the sameamount and has the same expected yield in real terms. The differences between A and Blie in the "timing" of their outcomes-not in the sense of calendar time but in the sense ofthe contingencies under which certain results appear. Thus asset A has a strong positivecovariance with the existing portfolio, while asset B has an equally strong negativecovariance. Rows 5 and 8 show the expected set of outcomes when we add assets A andB, respectively, to the old portfolio.

The bottom line of the covariance/portfolio analysis is given in rows 3, 6, and 9. Herethe various outcomes are translated into utility units, employing the utility functionshown at the bottom of the table. This reveals a significant difference between assets Aand B when taken in conjunction with the old portfolio and the degree of risk aversionimplied by the utility function. Thus while both A and B add an expected value of 100 tothe portfolio, A adds only 91 in terms of utility units, while B adds 100.8, measured in thesame terms.

The example of table 5-1 shows how risk considerations can lead to a significantdifference in the valuation of two alternative assets in the portfolio/utility-of-wealthcontext. How would we likely change this example if A and B were two projects beingevaluated from the social point of view? In the first place, the size of the projects wouldbe smaller. So, too, would the extent of variation in portfolio value as one moved fromstate to state. Furthermore, in all likelihood the "true" utility function would drop lesssharply in marginal utility as one proceeded through successive levels of wealth. Also,the variance of project outcomes would in all likelihood be far less. Finally, the

74 Arnold Harberger

correlation (positive and negative) between project outcomes and those associated withthis old portfolio would be far less than perfect.

Table 5-1. Asset Valuation under Different Risk AssumptionsExpected

State Depression Recession Normal Prospernty Boom Value Utility

1. Probability 0.1 0.2 0.4 0.2 0.1

2. Old portfolio 800.0 900.0 1,000.0 1,100.0 1,200.0 1,000.0 n.a.3. Utility

(Old portfolio) 790.0 897.0 1,000.0 1,099.0 1,194.0 n.a. 997.6

4. Asset A 0.0 50.0 100.0 150.0 200.0 100.0 n.a.5. Old portfolio + A 800.0 950.0 1,100.0 1,250.0 1,400.0 1,100.0 n.a.

6. Utility

(Old portfolio+ A) 790.0 949.0 1,099.0 1,240.0 1,372.0 n.a. 1,088.67. Asset B 200.0 150.0 100.0 50.0 0.0 100.0 n.a.8. Old portfolio + B 1,000.0 1,050.0 1,100.0 1,150.0 1,200.0 1,100.0 n.a.9. Utility 1,000.0 1,050.0 1,099.0 1,147.0 1,194.0 n.a. 1,098.4

n.a.: Not applicable.

Utility Function for Table 5-1

Wealth Utility AU Wealth Utility AU

800 790 n.a. 1,100 1,099 49850 844 54 1,150 1,147 48

900 897 53 1,200 1,194 47

950 949 52 1,250 1,240 46

1,000 1,000 51 1,300 1,285 451,050 1,050 50 1,350 1,329 44n.a. n.a. n.a. 1,400 1,372 43


Table 5-2 makes only two changes to the picture shown in table 5-1. First, the size ofthe new project (be it A' or B') is reduced from 100 to 20; that is, from 10 percent of initialwealth to 2 percent of initial wealth. Second, instead of the outcome ranging from zero totwice the expected value, we have the outcome ranging from 50 percent to 150 percent ofthe expected value. Asset A' still has a perfect positive correlation with the old portfolio,asset B' a perfect negative correlation. The utility function is still the same; in table 5-2 itis simply set out to read like an income tax table, to permit easy interpolation.

The change between the two tables is quite dramatic. Whereas the incrementalexpected utility from asset A was 91, and that from asset B was 100.8, we now have autility increase of 19.94 from asset A' and 20.30 from asset B'. The difference falls from 9.8all the way to 0.36-a much greater fall than one might expect just from reducing the


investment to 1/5 its former size and from cutting the coefficient of variation of theoutcomes in half.

One can also consider, using the data of table 5-2, how a project, C', whose outcomehad no cyclical sensitivity would compare with A' or 13'. If C' has an outcome of 20regardless of the state of the economy, the combination "old portfolio plus C' " wouldhave an expected utility of 1,017.60. Note that this is exactly 20 greater than the expectedutility of the old portfolio, 997.60. That is, in assessing the incremental benefit due to C',we get exactly the same answer, regardless of whether we work with expected income orexpected utility. In assessing that because of A', the expected utility measure differs fromthe expected income measure by 3/10 of 1 percent (19.94 versus 20.00). Finally for B' thedifference between these two measures is 1.5 percent (20.30 versus 20.00).

Now think of A', B', and C' as projects, and of the utility function. The project cost of20 represents 2 percent of national wealth. The cyclical swing of national wealth between"recession" and "prosperity" is equal to 20 percent of average wealth; the swing between"depression" and "boom" is 40 percent of national wealth. The probability of recession is20 percent; that of depression is 10 percent. In all these respects the scenario appears tooverstate the volatility of "states of nature" that we see in most real-world economies.Conclusion: Even in extreme examples we do not make much of a mistake by workingwith the simple and "utilitarian" assumption that society is neutral with respect to risk.This is what allows us to calculate an expected income, expected benefit, expected cost,and so forth in calculating the net present values of our projects.

On Discount Rates and Shadow Prices of Funds

This section reviews and extends the treatment given to this subject in "Reflections onSocial Project Evaluation" (Harberger 1985; pp. 169-75). I focus here explicitly on thediscount rate to be used in social project evaluation. We deal with a distorted capitalmarket in which we have a marginal productivity of capital, p (assumed = 12 percent), amarginal rate of time preference, r (assumed = 4 percent), and a market interest rate, i(assumed = 6 percent). We assume that when someone or some entity enters the capitalmarket with a new demand for funds, these funds are "sourced" from displacedinvestment (entailing a social cost of p per year) and from newly stimulated savings (witha supply price, hence social cost, of r percent per year). Thus we get a social opportunitycost of capital equal to COs (=fip + f2r), a weighted average of p and r, with the weightsrepresenting the fractions of sourcing from the two alternative sources. In our numericalexamples we will assumefl = .75 and f2 = .25, yielding cos =10 percent.

Some economists have argued for the use of r as the discount rate on the ground that ris the rate that individuals use in making their choices between present and futureconsumption. The use of such a low rate would seem to open the door to a whole flood ofprojects, but in fact it does not. These economists recognize the distortions existing in theeconomy, p and i differing mainly because of corporation and property taxes, i and rdiffering mainly because of personal income taxes. They recognize, in particular, thatdrawing 1,000 of funds from the economy entails (under our numerical assumptions)forgoing a perpetual stream of 90 per year (= 750 of displaced investment times itsproductivity rate of 12 percent) of productivity from forgone investment, and alsorequiring a payment of 10 per year (= 250 of newly stimulated savings times a supplyprice of 4 percent) of compensation to the new savers (required to elicit their choice to

76 Arnold Harberger

Table 5-2. Asset Valuation under Different Risk Assumptions and Project Size

Expected

State Depression Recession Normal Prosperity Boom Value Utility

1. Probability 0.1 0.2 0.4 0.2 0.1 n.a. n.a.

2. Old portfolio 800.0 900.0 1,000.0 1,100.0 1,200.0 1,000.0 n.a.

3. Utility

(Old portfolio) 790.0 897.0 1,000.0 1,099.0 1,194.0 n.a. 997.60

4. Asset A 10.0 15.0 20.0 25.0 30.0 20.0 n.a.

5. Old portfolio + A 810.0 915.0 1,020.0 1,125.0 1,230.0 1,020.0 n.a.

6. Utility

(Old portfolio + A) 8,001.4 912.9 1,020.0 1,123.5 1,222.2 n.a. 1,017.54

7. Asset B 30.0 25.0 20.0 15.0 10.0 20.0 n.a.

8. Old portfolio + B 830.0 925.0 1,020.0 1,115.0 1,210.0 1,020.0 n.a.

9. Utility 821.2 923.5 1,020.0 1,113.7 1,203.4 n.a. 1,017.90


Utility Functionfor Table 5-2

Wealth Utility

800 - 49 790 + 1.08 (W-800)

850 - 99 844 + 1.06 (W-850)

900 - 49 879 + 1.04 (W-900)

950 - 99 949 + 1.02 (W-950)

1,000 - 49 1,000 + 1.00 (W-1,000)

1,050 - 99 1,050 + 0.98 (W-1,050)

1,100- 149 1,099 + 0.96 (W-1,100)

1,150 - 199 1,147 + 0.94 (W-1,150)

1,200 - 249 1,194 + 0.92 (W-1,200)

save more). Thus we have a total flow of 100 per year, into the indefinite future, which isthe "cost" of raising 1,000 in the capital market today. Those who use r as the discountrate then proceed to discount this 100 per year flow at the rate r (here 4 percent) to obtaina present value of 2,500. From this operation comes a "shadow price of investible funds"of 2.5 (= os/r).

So we have a dichotomy between those who would use r as the discount rate andthose who would use os (=f1p + f2r). The first group is impelled to use a shadow price ofinvestible funds equal to (oS/r), while the second group uses a shadow price of 1.0 (thatis, makes no adjustment to the normal measure of capital outlays).

In my previous paper "Reflections," I mention three reasons for preferring thecombination (ct, 1.0) over the alternative combination [r, (os/r)]: communicability of theprocedure, implications with respect to current expenditures, and implications forhandling situations with different time preference for various groups. Here I will


concentrate on the second reason and refer readers back to "Reflections" for observationson the first and third.

It is but a simple step, following the numerical example presented above, to assumethat the 1,000 of borrowed funds was being used to pay for the salaries of nurses orpolicemen or tax collectors or teachers. If the shadow cost of the 1,000 is 2,500 when it isused to pay for electric generating capacity, that shadow cost must also be 2,500 when itis used to pay for the current expenses of government. Or, consider the quite realisticcase of a government that is always borrowing (debt increasing every year as theeconomy grows). Any reduction of 1,000 in current expenses-be it in nurses' or teachers'salaries, or in garbage or tax collection-will naturally lead to 1,000 less of borrowingthan would otherwise be the case, resulting in a saving of future costs having a presentvalue of 2,500. (This rather roundabout way of stating things is my response to those whosay that "governments don't typically borrow to cover current expenses." This maysometimes be true, but still any autonomous reduction in current outlays will have itseffect in reduced borrowing during the period in question.)

I am not arguing that the use of a 4 percent discount rate and a shadow price ofinvestible funds of 2.5 entails some fundamental flaw of logic or analysis. But I amarguing that the label "shadow price of investible funds" is misplaced. It should be the"shadow price of public funds," not just investible funds. Once this is recognized, onesees its powerful implication: Each and every current outlay should produce benefitsequal to 2.5 times its costs! (See Sjaastad and Wisecarver 1976.)

For myself, I prefer the much less shocking implications of using a)s as the discountrate and 1.0 as the shadow price of public funds (for both capital and current outlays).

Using Measures of Social Opportunity Cost

This topic is a general one, with applications to various types of social opportunity cost,but it is probably most relevant and also easiest to visualize in the case of foreignexchange. Using Em to represent the market exchange rate (expressed in real terms) andEs to represent the social rate, we have the following formula for Es:

Es= fi Em(1+ti) + XfjEm(1 + zj).

Here Em(1+ti) is the internal demand price of a dollar's worth of imports of type i, tibeing the tariff rate. (The dollar is used here to represent foreign currency-consider thatwe are a peso or rupee country.) Likewise Em (1+zj) is the internal supply price (marginalresource cost) of a dollar's worth of exports of type j; z1 being the subsidy rate applying toexports of type j. Clearly, fi and f; are weights giving the probable "sourcing" of a newdemand for foreign exchange from these various sources. (See tables in chapter 4 of thisvolume, "Project Evaluation for the Next Decade," for examples.) These weights mustnecessarily add up to one. Hence one can equally well express the social opportunity costof foreign exchange as

Es Em+X_fiEmti+YXfjEmz.i i

This is an important lesson. The social exchange rate can be looked upon as (a) aweighted average of internal demand prices of import goods and internal supply pricesof export goods or, equivalently, as (b) the market exchange rate (always in real terms)

78 Arnold Harberger

plus a weighted average of the distortions Emti and Emzj affecting the different types andclasses of imports and exports.

It is extremely important to realize that the social opportunity cost concept is simply away of systematically taking distortions into account for particular purposes. Itrepresents an application of the general proposition that, after making the "private"evaluation of a project or policy, one should then correct for existing distortions byapplying a general correction Zj Di AQi where Di is the distortion affecting the i' activityand AQi is the change in the level of that activity stemming from the project or policy inquestion.

The reason why we use the concept of social opportunity cost is based on therepetitiveness of the operations in question. The foreign exchange market does not knowwhether today's new demand for $5,000 will finance a pleasure trip, an insurancepremium, a computer, a lace tablecloth, or a shipment of children's toys. All the "market"senses is an extra demand. Its reaction will be the same, the "sourcing" of the extradollars will be the same, regardless of who comes in to buy the dollars or how the dollarsare finally going to be used.

One way to sense that this is the social opportunity cost of foreign exchange is toimagine a scenario in which somebody in a peso country goes out and actually buys$5,000 in the foreign exchange market. Then a fire occurs and the cash is burned up. Weknow that the private loss is $5,000, but what is society's loss? To replace the foreignexchange (which, in our terms, must be regarded as a real asset), the country will have toexport more and/or import less. On the reduction in imports, potential tariff revenue willbe lost; on the expansion of exports, any subsidies that have to be paid will add tosociety's cost, while any taxes that are generated will reduce that cost.

Return now to a normal case, where the $5,000 is not burned up but is spent on someimport good. The act of buying the foreign exchange is what defines the socialopportunity cost. Any distortions involved in the act of spending it have to be taken intoaccount separately. Thus if the $5,000 is spent on something with a 30 percent tariff, thisproduces a credit (in the social calculus) of $1,500; if the $5,000 is spent on somethingwith a 5 percent tariff, this credit is only $250. Whatever the case, the story of whathappens when the $5,000 is spent has nothing to do with the social opportunity cost offoreign exchange. It gives rise to a completely separate entry in the social evaluation of aproject.

The key word, in many senses, that defines social opportunity cost is sourcing. Themessage of this section is that sourcing and using foreign exchange are differentoperations, and the SOC of foreign exchange applies only to the former.

All the above refers to the case in which the project is going into the market to buyforeign exchange, which it then uses to purchase goods or services. There is also theopposite case of a project that, say, produces an export good and sells the resultingforeign exchange on the market. Here one just puts the previously described process intoreverse. What was the social opportunity cost of the foreign exchange bought by theproject in the earlier example now becomes the social benefit of the foreign exchange soldby the project. It is to the sale of foreign exchange that the weighted-average Es applies,and in this case it is a benefit, not a cost. By the same token, the tax or subsidy treatmentof the specific items that are exported by the project must be dealt with separately; eachseparate Di dQi should then be counted as one of the "extemal" costs or benefits of theproject.


Projecting Changes in Relative Prices

In "Project Evaluation for the Next Decade," I point out that there may be bigger room forerror in projecting market prices (relative to the general price level) than in estimating theappropriate relationship between social opportunity costs and their correspondingmarket prices. This naturally brings up the question of how to determine whenallowances should be made (in project evaluations) for expected future price changes.

Perhaps the first point to make is that the task of projecting the future time paths ofvarious important relative prices entering into project evaluations is not a task for theindividual project analyst. There is thus some reason to wonder whether this is subjectmatter to be "taught" in regular project evaluation courses. What definitely should betaught is that project evaluators should be alert to the dangers of assuming that relativeprices will simply remain constant into the indefinite future. They should seek outinformation on prospective changes and should incorporate such changes in theirevaluations of particular projects.

In the best of all worlds, there would be central bodies to whom project analysts couldturn in getting the required information. It is my understanding that the World Bank hasa unit whose task it is to assess likely future trends in the prices of the principalcommodities involved in world trade. It would be extremely helpful if this unit couldtake as part of its task the production of publications that could be used almost as"manuals" by project evaluators all over the world. Such a unit might also develop (aloneor in collaboration with the International Monetary Fund) a general index of the "worldprice level of traded goods," which evaluators use as the standard deflator of the worldprices of specific commodities (and could also use as the external price level in thestandard measurement of each country's real exchange rate).

Even if the World Bank or some other organization undertakes the task of providingguidelines on relative price changes of internationally traded goods, there still remains asignificant role for the central project evaluation authorities in individual countries. Notonly should they make their own studies of the particular products that are mostimportant in their country, but also they have major macroeconomic tasks, such asprojecting the short- and longer-term future movements of the real exchange rate, thesocial opportunity cost of capital, the overall average level of real wages, and any likelytrends of the wages of particular categories of workers relative to the overall averagelevel.

Let us examine a few extreme cases. Today the consensus is that Argentina's realexchange rate (measured in pesos per dollar) is too low. The American press recentlyreported that Buenos Aires was the most expensive city in the Western Hemisphere fortraveling businessmen. Even Argentina's official representatives recognize that today'sreal exchange rate is not tenable in the long run. (Their hope is that internal productioncosts will decline in real terms, so as to render a devaluation of the nominal pesounnecessary.)

In the early 1980s a spate of currency overvaluation occurred in a set of countries(Argentina, Chile, Mexico, Uruguay, plus others) that were the recipients of vast inflowsof foreign capital. These inflows made the dollar extremely cheap (to local buyers) in realterms. But it could easily be verified that the rate of capital inflow was too high to besustainable. The obvious conclusion was that the long-term real value of the dollar wassignificantly higher than the prevailing market value then.

80 Arnold Harberger

These are real cases in which the evidence favoring a future rise in the real exchangerate was extremely strong. All four of the listed nations soon became the victims of theinternational debt crisis. Huge demands were made on them, mainly by the internationalbanks that were their creditors, and their economies were put through wrenchingadjustments. In the process their currency underwent huge, real devaluations. In each ofthe mentioned cases, the monthly series on real exchange rates has a peak that is morethan twice the preceding trough. Moreover, this great swing took place within arelatively short time after the debt crisis struck.

Just as it was clear that the prevailing real exchange rate was below its long-run normwhen the flood of capital was coming in, so too in the height of the adjustment thatfollowed the debt crisis it reached points that were well above the norm. It is importantfor readers to realize that I am not asserting that these were points of disequilibrium-itis better to think of them as points of equilibrium characterizing transitory circumstances.When the capital flow was large, the supply curve of foreign exchange could be thoughtof as shifting far to the right. This was transitory because the countries in question weremoving quickly toward their maximum debt/GDP ratios. The existence of prudent limitsto the debt ratio can hardly be denied, but continued borrowing at the pace of 1980 or1981, depending on the country, would soon carry it beyond such a limit.

In reaction to the debt crisis the countries squeezed credit to the extreme as theystrove to meet the demands for payment. In the process GDP fell by an average of morethan 10 percent in the four countries; imports were cut by more than half; unemploymentsoared. All these came as the result of shifting the countries' stance from one of receivinga massive inflow of capital to one of having to generate a huge trade surplus. The realexchange rate had to be high in such circumstances in order to deter imports andstimulate exports. But the short-run "equilibrium" of that moment was different from thelonger-run equilibrium that would come later, if for no other reason than the differenceof supply response between the short run and the long run. When the debt crisis struck,the great bulk of the adjustment was borne by imports. Later, in response to the higherreal exchange rate, exports (particularly nontraditional exports) grew significantly.

It should now be clear why the real exchange rate was low (but in equilibrium, givena huge capital inflow) in the early stage and high (but in equilibrium, given the demandsof creditors) at a later stage, and why the longer-run equilibrium reflecting a "normal"postdebt-crisis situation would produce a real exchange rate that was in between, butquite distant from either of these extremes. Hence a project analyst working at the time ofthe big inflow should properly assume that the real exchange rate would increase fromits then-prevailing level, while an analyst working at the time of the debt crisis shouldwith equal assurance project a decline in the real exchange rate (RER).

Other circumstances in which a change in RER should be projected include caseswhere a country's principal export good is experiencing an unsustainably high or anunsustainably low international price. In these cases there would be two price trajectoriesto trace-one for the principal export product as such, the other for the real exchangerate. Examples include coffee in 1977 (its world price tripled between 1975 and 1977),sugar in 1974 and 1980, tin in 1978-79, rubber in 1980, cocoa in 1977-78.

In any of these cases one has to study the situation to make sure that the new pricedoes not reflect a permanent change in the prevailing supply/demand situation-something that is generally quite feasible to do. (Sugar reached 40-50 cents a pound in


1974 and 1980. At those prices there is hardly an arable hectare on earth where sugar, ascane or beets, cannot be profitably grown.)

The social opportunity cost of capital should be connected to the measuredproductivity of capital in the economy, but in periods of clear stringency in supply offunds in relation to demand, the near-term rates of discount should rise above the longer-term rate. For several years Chile used a variable real rate that started at 20 percent forthe first year, then went to 18 percent for the second, 16 percent for the third, and so on,down to 12 percent for the fifth year and beyond. This is an exemplary way to deal with anear-term stringency of funds that is not expected to be permanent.

The course of real wages will largely be determined by the growth rate of GDP perworker, so a country projecting a 6 percent growth rate of real GDP will have a muchdifferent outlook for real wages than one which projects GDP growth at 3 percent peryear. GDP growth is the principal factor to consider when projecting the growth of thegeneral level of real wages, but we must also look at the situation of different segments ofthe labor market. When developing countries pass through a stage of rapidmodernization, liberalization of capital and financial markets, and so forth, significantlabor market premiums typically arise for the particular skills associated with financialanalysis and the management of financial institutions. At the same time, modernizationdoes not exert pressing demands on the unskilled labor force, so it is common for theunskilled wage rate to lag well behind the average rise of real wages duringmodernization periods.

The preceding notes reflect the level at which I would attempt to communicate someof the issues surrounding the projection of relative prices, in a course of relatively shortduration, devoted to the general topic of social project evaluation. They do not go farenough to be the guidelines for those who will actually do the projections of relativeprice trajectories. But they certainly are adequate for sensitizing participants to the needto build changing relative prices into their projection procedures. They also indicate someof the main types of situations in which this need is most likely to arise.

References

Harberger, Arnold C. 1985. "Reflections on Social Project Evaluation." In Pioneers in Development.Second Series. Washington, C.D.: World Bank.

Sjaastad, Larry A., and Daniel Wisecarver. 1976. "On the Opportunity Cost of Public Finance."Journal of Political Economy.

6

The Role of EDI in Project TrainingforDeveloping-Country Managers

William Ward

The Economic Development Institute (EDI) should take a "worldwide systems" view ofits role in project training in which it looks at national, regional, and Washington-basedactivities in terms of helping developing-country organizations create policies, operatingprocedures, and skills suited to development thinking and the development problemsthey will encounter in the 1990s. Pursuant to this, it should address the following issues:

1. Project analysis training often does notfocus on the right kinds of projects. Project trainingprovided for government officials in developing countries tends to be oriented towardplanning and appraising projects (generally, larger ones) funded by bilateral andmultilateral donors. The donors do their own appraisal; thus national agency staff needsto provide a liaison function, without necessarily being able to do these kinds ofappraisals. Central and line agencies also need training directed toward projects that arelocally planned, appraised, and funded. At least the national training programs shouldfocus on providing this type of training. Moreover, the bulk of this training should betargeted toward the standard operating procedures (SOP) of the organizations doing theproject work. (See point 2, below.) Thus the kinds of project skills taught shouldcorrespond to the functions of the various officials. This issue will be the focus of furtherdiscussions on the venue of training.

2. Policy and procedures (management) consultancy often should precede project analysistraining. National agencies should have in place organizations, policies, and procedures(guidelines, manuals, and so forth) around which to develop project training. Beforecountries can develop the right types of training programs, they need managementconsultancy to help put into place the appropriate organizations, policies, and procedures.Time will often be better spent in the early stages by assisting in the development ofthese parts of the system as a first step. Is this kind of needs assessment/managementconsultancy an appropriate role for the EDI?

3. We need to do a better job in dealing with issues other thanfinancial and economic analysis.Project courses and materials have focused on financial and economic appraisal ofprojects because these are easiest to teach in a short course format. Major weaknessescontinue, however, in other areas of project planning and appraisal. For example,developing countries continue to have great difficulty finding the specialized subsectorexpertise needed to conduct technical analysis of industrial project proposals. Finding

83

84 William Ward

some mechanism for putting technical, marketing, and operation and maintenanceexpertise into project planning continues to plague project trainers. This involvesrecruitment and retention practices, as well as training. What role do the Bank and theEDI have here?

4. We need to rebalance thefocus within economic analysis. Project economic analysis has,since the publication of the Manual of Industrial Project Analysis in Developing Countries byLittle and Mirrlees in 1968, focused on government failure and corrections for pricedistortions in the analysis. Much less attention has been paid in the past twenty years tothe issues of identifying and valuing the effects of projects directed toward dealing withmarket failures, classic public and quasi-public goods, externalities, and so on. Issuessuch as estimation of willingness to pay for public utility outputs (that is, the consumers'surplus issue) and valuation of quasi-public goods and externalities have taken a backseat to issues that include the estimation of conversion factors and accounting prices. It isprimarily in industry and agriculture where the price distortions issue has warpedinvestment decisions. Yet most of us have spent undue amounts of time in the pasttwenty years addressing these issues in all sectors. The EDI's role in developing materialson water supply has been beneficial in breaking this trend. Can we muster the resourcesto continue filling such voids?

5. We need to better integrate project analysis and policy analysis as well as the relatedtraining. Economists now widely agree that we will not likely be able to make enoughproject investments in any sector in the 1990s to make up for bad policies affecting thesector. Thus in most countries the focus has shifted to getting policies and prices rightrather than to correcting for price distortions while going forward with a government-directed investment program. The donors have moved on to implementing this approachin their programs without gaining an understanding of these issues by most governmentofficials who are being asked to implement the new approach. On the one hand, we areteaching project courses based on an old government-interventionist, accounting-pricemodel while, on the other hand, pushing programs that involve policy reform. Amongother things, the content of the policy workshops needs to be brought into the projecttraining programs. In addition, the nature of project analysis training must reflect thechanging policy environment within which projects are planned. Do we need to do abetter job of linking the policy workshops and the project training program?

6. We need to integrate recent changes in development theory into project analysis and relatedtraining. World Development Report 1991 reflected a changing approach to development.Compared with the 1960s and 1970s, market failure justifications for public sectorintervention have narrowed greatly. These were narrowed further by the suggestion thatgovernment failure arising from inappropriate public sector interventions frequentlycauses more problems than simply tolerating the inefficiencies posed by many categoriesof acknowledged market failure. We need to do much more in all the project courses toreflect the changing views on which sets of government activities are now deemedappropriate and supportable by the intemational donor community. We also need to domore to develop appraisal and training materials that integrate this new approach intoproject appraisal practice, both inside and outside the Bank. At this juncture, the Bankhas not accomplished this integration even for its own methods of project appraisal. Thisis partly because few people in the Bank are paying much attention to continuing toimprove project appraisal practices (see point 7, below). Can the EDI take a leadershiprole in improving appraisal practices beyond those currently being applied in the Bank?

Role of EDI in Project Training 85

7. Bank practice cannot be the ultimate model for project analysis training in the EDI. Since1980, project analysis work in the Bank has become increasingly perfunctory. Inoperational terms, lower-level staff are involved in project appraisal; fewer resources arecommitted to appraisal; basic models are seldom being developed for creating overallbenefit streams; and less oversight is given to the appraisal process. On the methodologyside, the Bank is doing less work on development and refinement of methods for projectanalysis (with the possible exception of environmental impact analysis). The Bank'soperational experience is becoming less of a source of practice and materials for projectanalysis, per se.

8. It will take special efforts to make Washington-developed training materials relevant tonational programs. Certain training materials used in the national programs will needspecial attention, especially those related to procedures. Generic materials developed inWashington will not meet the targeted needs, although there will be a need for somegeneric materials. In addition, case materials will continue to be problematic because ofthe scarcity of case-writing skills. Many national programs have no staff with case-writing capability. Can the EDI provide a corps of case writers to work with nationalplanners to help them develop appropriate case materials?

The Evolution of Project Training in the EDI

From the mid-1960s to the early 1980s, the EDI was a direct provider of project trainingfor officials from developing countries. (The term "project training" is used here todenote training in the areas of project identification, preparation, appraisal, supervision,monitoring, and evaluation.) Training was conducted in three forums: (1) Washington atthe EDI, (2) regional programs at various multinational institutions (for example, forMiddle Eastern officials at the Arab Planning Institute in Kuwait), and (3) nationalprograms in developing countries. In addition, the EDI assisted universities and otherinstitutions in both developed and developing countries to prepare materials andcurricula for project training within their graduate degree programs (for example, ShirazUniversity, the University of Pittsburgh, Michigan State University, Bradford University,and the University of the Philippines).

EDI project training drew on the Bank's operational experience, and new coursedevelopment depended heavily on the Bank's operational staff. It conducted training-needs identification in conjunction with operational staff, who also had an important rolein determining course content and selecting training materials. The typical pattern was toput a large initial trench of EDI and Bank resources into developing a course forpresentation in Washington, before using the new curriculum and materials overseas innational and regional project training programs. The bulk of the materials now availablein the EDI for project training was developed in conjunction with Washington-basedcourses and revised in repeated Washington and overseas presentations.

The Washington forum provided the developing and testing grounds for many of thecourses that in the early 1990s continue to be conducted in revised form overseas. TheEDI presented some project courses in Washington only once (for example, the 1977General Projects Course); others were repeated several times there (for example, theIndustrial Projects Courses). Because the EDI tended to allocate more resources to theWashington courses, those most often repeated there had and continue to have a larger,

86 William Ward

better developed body of training materials available (see EDI Catalogue of TrainingMaterials).

By the mid-1970s a stated objective of the EDI was to work toward transferring theproject training capacity to regional and national institutions in the developing countries.By 1982 the institute had given primary responsibility for conducting project trainingover to these institutions.

EDI's staff worked in the late 1970s to identify suitable regional and nationalinstitutions to which to transfer this capacity. Almost no perfect-match institutions werefound. In the developing countries at that time, training institutions trained governmentofficials in administration, national development planning, and various other subjectsand skill areas; however, there were few national or regional institutions with a focus orcapability in project training. In many cases, the EDI had to exert some degree ofinfluence to get the institutions to take on project training as an area of work.

Various institutions were tried at different stages in the process, includingadministrative staff colleges, such as the one at Lahore; colleges, academies, anduniversities, such as the University of the Philippines at Los Banos and the PakistanAcademy for Rural Development; regional management institutes, such as the Easternand Southern Africa Management Institute (ESAMI), training departments ofdevelopment banks, including the Private Development Corporation of the Philippines(PDCP). In each case, course directors and lecturers had to be identified and trained tocarry out the project training task and new career paths had to be built. In addition, theinstitutions required assistance in developing the training curricula and materials.

Differences between Project Courses in Washington and Overseas

Staff levels and training venue. Early in the process of transferring the capacity to nationalinstitutions, it became apparent that important differences existed between Washington-based and national and regional training programs. First of all, the Washington programstypically drew higher-level, faster-track officials, most of whom shortly becamemanagers-not only managers of the project's function itself, but often in a very shorttime policymakers and senior managers. The national programs, in contrast, trained afew (primarily midlevel) managers in the early presentations and then focused largely ontraining the more pedestrian staff of the projects' agencies. This pattern was consistentwith the real-world interface between training and compensation that existed then andhas continued in most public institutions in developing countries: Because of low salariesin many of the subject institutions, overseas training has always been considered aspartly staff development and partly compensation, with the more attractive overseasvenues used to compensate higher-level staff for the financial sacrifices associated withremaining in the public service.

Content of Washington programs. The Washington project courses stemmed largelyfrom Bank projects and procedures. They taught the lessons from a fairly wide range oflarger projects, and they taught procedures that had been fairly well developed by Bankpractitioners for the organization's needs and projects. The Washington courses werevery good training for those government officials who worked on Bank loans or hadother liaisons with it. For the most part, the cases and procedures developed inWashington were revised and taught in the national and regional programs-although


there were and continue to be many examples of materials developed specifically for andfrom the overseas courses.

Project work in national agencies. Most of the larger projects in developing countriestend to get picked up for funding by the bilateral and multilateral donors. The donorshave their own planning and appraisal procedures, and donor staff or consultantsprepare and conduct the appraisals. The national line and central agency staff conserveresources by not duplicating this work. Rather, the theory is that national staff providethe link between the national development plan and the donor agency financing plansfor the national investment program. Project planning and appraisal skills directedtoward more than the requirements for this liaison function for the larger projects maywell represent wasted resources.

Local projects and national training programs. What often has not been adequate in thenational programs is development of training and materials oriented toward planningand appraisal of the smaller and more numerous projects funded outside theinternational donor framework. Most of these projects could benefit from improvementsin planning and appraisal; the smaller projects normally found in the national programscannot, however, support the cost of the kind of appraisals conducted for the largerprojects funded by donors. Developing countries need appraisal methods suited to thetypes of projects and appraisers found in their agencies. Thus far, their training materialsand programs have lagged behind those developed for planning and appraising thelarger projects.

Project organizations and project planning and appraisal. In the performance-basedtraining model discussed later in the Appendix, project training in at least the nationalprograms would correspond to the organizational structures, planning and appraisalprocedures, techniques, and manual statements that apply to the working environmentof the officials in the countries' organizations. In most cases, the national trainingprograms developed thus far have not given due consideration to the context of thisoperational environment. In the face of these failures, many countries have not "boughtinto" project analysis procedures and skills for officials planning local projects, andconsequently have left the application of project appraisal techniques to the donors, sincethe donors will do their own analyses anyway.

National, regional, and Washington programs. The audiences and needs of the threevenues are different. The audience of the national programs had and largely continues tohave at least two characteristics not well served by the materials and approachesdeveloped in Washington: (a) Most of the officials being trained in the national programsare low- and mid-level technicians, many of whom have not studied project analysis evenin abstract form and lack even the low-level skills associated with acceptable projectplanning; and (b) in many cases, the officials are still planning and appraising projects inthe absence of national equivalents of operational directives, operational manualstatements, or central projects notes (that is, SOPs). The audience of the Washingtonprograms tends to be higher-level professionals and managers who need to focus moreon policy and management issues. The regional programs tend to fall in between, with amix of audience ranging from staff similar to those of the national programs to thehigher-level staff attracted by the Washington programs.

88 William Ward

Venue, Curricula, and Materials

Training venue, curricula, and materials should tie in with the objectives of the trainingactivity. The model discussed in the Appendix presents one "appropriate" approach torelating categories of objectives to categories of venue. It is important also to relatetraining materials development to objectives and venue.

In general, it is more feasible to develop abstract materials outside the venue of usethan to develop concrete materials away from the venue of use. The EDI's experience inthe 1970s, for example, was that project analysis case studies tended to grow out of aparticular trainer's need to demonstrate a concrete point (or the intersection between anabstract principle and a concrete lesson) in a session that he or she was directing. Seldomdid usable cases derive from any other source than a trainer's need for the material for aparticular session. (Typically, the cases then were picked up and used by other trainersafter seeing the case taught two or more times.) A small number of lecturers in the EDIdeveloped most of the materials of this type specifically for project courses being offeredrepeatedly, either in Washington or overseas. This basic observation holds, in particular,for the development and revision of case and exercise materials.

Training materials relating to general principles or to reviews of knowledge andexperience inside and outside the Bank have been developed by a broad range of partiesfor use inside and outside the EDI. These have not necessarily been related to the specificin-class needs of trainers. The experience with materials prepared by consultants, forexample, has tended to be much better in course note writing than in case writing. Byextension, the range of uses (in terms of training venue) of the course notes has tended tobe better than that for the cases-although we must recognize that the lack of good,specific cases has often led to usage of inappropriate cases in a wide range of venues.

The hypothesis arising from these observations is that it will likely be difficult toprepare skills-oriented training materials in Washington for use in national and regionalproject training programs unless two requirements are met: (1) Those who prepare thematerials are the ones presenting the sessions in the overseas courses, and (2) the EDIallocates sufficient time and resources for these lecturers to identify, prepare, test, andrevise the materials. This hypothesis must be tempered by an additional observation:Case preparation skills (perhaps like fiction writing skills) are scarce. Among all thelecturers in the EDI's history, probably less than 10 percent have proved willing and ableto write usable case studies. Thus the preparation of usable case materials not only willpose the above two requirements, but also will require that adept case writers be identifiedand assigned overseas teaching responsibilities, coupled with the time and resourcesrequired to develop such materials.

Venue and "skills" orientation. The national programs have always needed more"skills" orientation (that is, oriented more toward procedures than policy) and moretargeting than the university programs, the regional programs, or the Washington-basedEDI programs. They continue to have a substantial in-country demand for country-specific and organization-specific skills in project planning, analysis, and management.Materials and curricula-especially for these programs-tend to be most appropriate ifdeveloped from local needs, cases, procedures, and conditions.


A Model for Relating Venue and Training Content

Established procedures and the performance-based training model. A major problem with thenational training programs in many countries is that established and codified policies,procedures, and techniques do not exist (with differing degrees of exception, such as thePhilippines and Pakistan. They seldom have official procedures (such as the Pakistan PC-1 form and related systems and the Philippine "Project Development Manual") on whichto design a project training program based on the classical performance-based model.This is particularly the case in the countries most frequently thought of as having serioustraining needs (largely the poorest member countries, many of which are in Africa). Thusto meet the needs at the national level, in many if not most developing countries, maywell require a substantial commitment of resources in the form of skilled and specializedmanpower in organizational development and training in project institutions.

The procedures/training interface. Most of these countries need to establish national andorganizational systems first before they can effectively put in place the type of trainingnecessary. (This approach was effectively used, for example, in the Philippines in theearly 1980s for a regional planning project of the United Nations DevelopmentProgramme implemented by the Bank, under the leadership of former Bank divisionchief Ben Thoolen.) Thus one may argue that the "off-the-shelf" approach to curriculaand materials is substantially less than satisfactory for those agencies that needorganizational development the most. On this basis, one may question how a modelusing training materials developed in Washington and made available to national andregional training programs could help the countries and institutions needing the greatestassistance.

A training model. The Appendix to this report presents an approach to project trainingbased on what might be called a "classical" model. It puts training into the broadercontexts of (1) strategic planning and management, (2) organizational development toachieve the performance objectives stated in the strategic plan, and (3) training as just onesubset within strategic planning and organizational development. This approachinvolves the trainer in the overall planning activities of the organization, and it involvessenior management in the identification and definition of the training program. Themodel breaks the distinction between "staff" and "line" functions and attempts to bringtraining (organizational development) back into the line functions. The approach is incounterpoint to the typical pattern found in many organizations, where managementtypically delegates training to a "staff" functionary (who often further delegates to aconsultant) and does not involve trainers in the "line" functions of the organization. Thislatter approach, although widely practiced, has proven repeatedly to be a model fororganizational performance failure.

The proposed model has certain attributes that trainers should appreciate at theoutset. First, training and management are parallel and intertwined. Training materialsand curriculum preparation-in practice-must correlate with the development oforganizational policies and procedures. The development of materials and curriculaneeds feedback from, and interaction with, managers to make the materials good;furthermore, management needs the interplay with the training function in order todevelop good policies and, especially, procedures.

Second, training and organizational development are continuous and dynamic. Onlyin a static organization, in which most activities are routine, would training curricula and

90 William Ward

materials stay the same for long. If the environment, objectives, or policies of theorganization change, the training also needs to change. Thus trainers must be constantlyinvolved with line managers and operational staff. The changing environment indeveloping countries posed by structural adjustment, for example, provides a case inpoint.

Policy seminars and dynamism in procedural functions. Policy-oriented training andeducation should be designed to prepare upper management to deal with nonroutineactivities and to make policy. Procedures-oriented training should be developed inconjunction with and reflect the policies and procedures that evolve from strategic andtactical decisions made by upper management. If there is a need for policy training, then,by definition, there will also be a dynamic element to procedures and to the training thataccompanies it. It is not rational to expect that project training curricula and materialswill be static in the face of a perceived need to conduct policy-oriented seminars.

New policies, old proceduires. The 1980s saw a dramatic change in the environment inwhich project work takes place. The materials arising from the various EDI policyseminars document this change quite well. The EDI spent the decade-no doubtcorrectly-attempting to help senior managers cope with that changing environment. Inthe process, it paid less attention to the interface between policies and the proceduresneeded to implement changed policy approaches. This is true across a broad spectrum ofdevelopment management functions in developing countries-including macroeconomicmanagement and program and project planning. Although the EDI was helpingmanagers understand the changing environment, the "procedures" used by lower-levelstaff in the organizations reflected policies designed to deal with a paradigm that grewout of the "policy" issues of the 1960s and 1970s. It is time for the EDI to consider (1)helping convert the evolving policy changes into SOPs that will improve the performanceof the organizations whose managers have been sensitized, and (2) training the middleand lower-level staff, who must now operationalize these performance-based changes inSOPs.

The Current State and Changing Requirements of Project Analysis

Project planning and analysis require a broad range of skills. These fall under the generalheadings normally assigned to "Aspects of Project Analysis"-namely, financial,economic, commercial, technical, organization and management, and so on. In the 1960sand 1970s, the EDI attempted to develop training materials and approaches to deal withall the aspects of project planning and appraisal. Staff assigned to project trainingincluded engineers, agriculturalists, financial analysts, and even management experts.

By the end of the 1970s, the project courses had settled down to focusing largely onfinancial and economic aspects of projects. The other skill areas were still thought to beimportant, but most lecturers tacitly had begun to admit that (1) not all those objectivescould be met in one training program, and (2) certain of those aspects could not betransferred in the time available, even if the entire course were allocated to it (forexample, market planning and analysis, organization and management, and technicalanalysis).

Most of the project analysis material and curricula currently available within the EDIfocuses on financial and economic analysis, much of which was developed for


approaches to development planning and project analysis in the 1970s. Besides theproblems of dealing just with financial and economic analysis, these latter materials arein many respects even inconsistent with new approaches to development and withproject appraisal needs in the 1990s-a problem shared with the Bank's old OperationalManual Statements and Central Projects Notes (CPNs). In other words, not only will theEDI have to rewrite much of its project appraisal material to accommodate these changes,but also the Bank's new operational directives will need to reflect the changes.

Theories of development practiced in the 1960s and 1970s gave a proactive role togovernment. The presumed success of central planning in the former U.S.S.R. and thebelief that widespread "structural" problems in developing countries precluded theefficient functioning of markets led to prescriptions for large-scale governmentintervention and "directing" their economies. Projects were often deemed the cuttingedge of this process.

The activist role of government in the 1960s and 1970s led to large-scale "distortions"in prices and incentives in many developing countries. The effect of these distortions onpatterns of public and private investment led project economists to turn to dealing withthese distortions in the analysis of public sector investments and publicly influencedprivate investments. This process came to a head in the late 1960s and early 1970s withthe publication of two important, widely known books on project analysis by theOrganization for Economic Cooperation and Development and the United NationsIndustrial Development Organization. These books and their descendents caused cost-benefit analysts to shift their attention almost entirely from issues related to dealing with"nmarket failure" in project analysis to dealing with "government failure." Infatuationwith conversion factors and accounting prices was the outcome of this shift in focus ofcost-benefit analysis.

In the 1980s the focus was on getting government out of economic activity rather thanone of adjusting the project accounts for government's distorting effects. Instead ofsuggesting investing in projects that looked good after correcting for the distortions-theapproach taken in the 1970s-analysts shifted to "sector" and "policy" operationsdesigned to get rid of the distorting influences imposed by government. In the process,the pendulum may have temporarily swung too far, and-with encouragement from theThatcher and Reagan administrations-it became popular in the Bank and other parts ofthe donor community to suggest that all government activity may be inappropriate.

By 1990 a more balanced approach had begun to re -emerge in the Bank and in someother parts of the donor community. The more balanced approach returned to anupdated version of the "theory of market failure," in which government was given a role,but a more circumscribed one. World Development Report 1991 recognized this changedworld and a new approach to development involving a facilitating role for government,rather than a proactive role.

This reorientation of development theory is already having an inexorable effect onproject identification, planning, and appraisal. On the one hand, it is leading toimplementation of new types of projects (for example, the India TechnologyDevelopment Project) by those who are at the forefront of the emerging approach todevelopment. On the other hand, it is causing analysts to apply the traditionalapproaches to project appraisal-those reflected in existing manuals and guidelines- inan increasingly perfunctory fashion, both in the World Bank and in agencies that stillpretend to do cost-benefit analysis. The guidelines and training materials that form the

92 William Ward

basis for "procedures" of project analysis are increasingly inappropriate to thedevelopment theories and project practices of the 1990s.

Dealing with project-planning needs in the 1990s will require a major refocus oftraining materials and curriculum development:

1. Within the proposed worldwide delivery system we must find some way toprovide the skills for other aspects of project planning besides financial and economicanalysis so that the line agencies will have project planners skilled in these other areas.

2. Curricula and materials must be developed to assist government officials inidentifying and planning the public interventions deemed appropriate in the new andevolving approach to development-for example, skills in identifying interventions toimprove the environment for developing and internalizing technology.

3. A refocus of cost-benefit analysis on the identification and measurement of benefitsarising from interventions in the market failure sectors is necessary and must lead tocurricula and training materials directed toward developing the requisite skills.

4. In the next five years, training materials and curricula will be required to helpestablish this new project paradigm in the consciousness of central and line agencymanagers and professionals.

These changes will be particularly problematic. It takes ten years or more to get "new"thinking in the economics community to effect the policies and procedures of operatingagencies in developing countries. First, the attitudes of top management must change.Then, policies must change. Finally, training of operational staff and the development ofnew procedures must proceed together.

We currently find ourselves running training activities designed to reorient thethinking of top managers to the "new" approaches to development. Bank operationalunits are, at the same time, trying to force changes in policies to accommodate the "new"thinking. But at the same time training programs and project-planning procedures stillfocus on "old" methods of public intervention in the economy.

In the 1990s, major new investments will be necessary in the development of"systems" for implementing what we learned in the 1980s about the role of the publicsector and about the role and practice of project planning and implementation ineconomic development. Just as the Bank, through the EDI, took a leadership role in theseareas from the 1950s onward, a similar leadership role will be essential in the 1990s.


Appendix: A Systems View of Human Capital Development for Project PlanningOrganizations in Developing Countries

The differences in training needs of various officials and in the strengths and weaknessesof different venues lead to identification of a model based on a worldwide systemsapproach to meeting the human capital needs of project planning and implementationorganizations in developing countries. The model involves at least two sets of actors: (a)the individuals and institutions to whom the training and education are directed, and (b)the actors in the training and education delivery process ("training" implies here deliveryof concrete skills, while "education" implies delivery of broader and perhaps moreabstract knowledge).

Project Training Recipients

There are different uses and thus different needs for project training at various levelswithin project organizations and central organizations of developing-countrygoverrnments. Staff members passing through the management progression need, at sometime or another in their careers, an understanding of the training and educational contentdelivered at each position level they enter.

Generally, officials at higher levels (policymaking and management positions) needgreater breadth of knowledge of the political and economic environment and the policyissues surrounding project planning. They require more knowledge of the externalenvironment within which the organization operates. They also need to know somethingabout the general principles that have been developed in the broader world to deal withthe issues involving their organization.

Staff at lower-level operational positions in project organizations, in contrast, needgreater depth of skills in the details of the organization's work. The requirements for thesepositions tend to be (a) more internally focused and (b) more skill oriented than are thesenior management and senior policy positions. The lower-level staffs knowledge andskills are usually quite specific to the procedures and practices of their ownorganizations. Table 6-1 presents a schematic view of this distinction. This difference issometimes referred to as the hierarchy of emphasis within the organizational structure on"policy" issues compared with "procedures" issues. Note in table 6-1 that the high-levelofficials do not necessarily have to be as knowledgeable about the details of internalprocedures as do the midlevel managers and the senior professionals. However, the mostsenior officials should be much more informed about the environment in which theorganization functions. This latter kind of knowledge often will require continuousexperiences outside the organization, even outside the country.

In general, training tends to be increasingly specific, skills-oriented, and proceduralwhen the training venue is closer to the job site. Many organizational developmentexperts argue that it is best to address the needs of lower-level officials in organization-level training activities, those of midlevel officials in national training programs, and soforth. Senior managers and senior policy advisers need a much broader view of theexternal environment in which their organizations function; thus the policy-orientedseminars offered by the EDI for senior officials tend to fit well into this systems approachto training.

94 William Ward

Table 6-1. External Knowledge Compared with Internal Skills Requirements of Officials in ProjectPlanning OrZanizations

External Internalknowledge skills

Official (policy) (procedures)

Senior managementSenior advisersUpper middle managementMiddle managementSenior professionalsProfessionals **

Assistant level *

Project Training Delivery

As implied in the preceding subsection, project (as well as other forms of) training maytake place close to the job, or at great distance. At the same time, the training may beoriented toward organizational procedures (that is, it may be very concrete), or it mayfocus on principles (it may be more abstract in approach). Table 6-2 schematically statesthe principle that the breadth of vision required by senior management may require thatpart of their education and training be done in programs in which they interact with abroad range of peers from other related environments. Lower-level staff are engaged inmore "procedural" activities (see the following section and table 6-3). Generally, thelower the operational part of a project organization, the more procedural the trainingneeds tend to be. At the lowest levels (that is, assistant level and professional level, intable 6-2), knowledge and skills in established procedures diminish because: (1) newerentrants to the organization tend to cluster at those levels, and (2) older staff with limitedability to master all skills required for the organization's work also tend to remain atthese levels.

Routine and Nonroutine Project Activities

Generally, the most senior managers in a project organization should deal almost whollywith nonroutine activities-tasks for which no established procedures are possible andthus carry the possibility of making policy with every decision. Standard models oforganizational management suggest that the mix of routine and nonroutine activitiesshould change toward more routine as one moves downward in the organizationalhierarchy. Table 6-3 describes the suggested relationship between routine and nonroutineactivities according to staff level within a project organization.

Generally, the more routine the activities performed, the greater the scope fordeveloping established procedures, manual statements, guidelines, and classical training.The more nonroutine the activity mix, the greater the need for abstract thinking,knowledge of general principles and policies of the organization, and knowledge of theoutside environment. Again, the EDI policy seminars generally help build the capability

Role of ED] in Project Training 95

Table 6-2. Training and Education Venuesfor Human Capital Requirements of ProjectOrjanizations

External Internalknowledge skills

Venue (policy) (procedures)

International institutions

(for example, EDI, Harvard University)

Foreign regional institutions

(for example, Arab Planning Institute,

ESAMI, etc.) ****

National university ******* **

Local or regional college

Staff training college

(for example, administrative staffcollege)

Sectoral training institute

(for example, LPPI, bankers' stafftraining college, and so on)Training department

On-job supervision

Table 6-3. Routine and Nonroutine Activities by Level in Project OrjZanizations

Nonroutine Routineactivities activities

Official (policy) (procedures)

Senior management *

Senior advisers

Upper middle management

Middle management

Senior professionals

Professionals

Assistant level *

to deal with nonroutine activities, and to establish the policies for a framework ofprocedures.

The Procedures/Training Interface

Not all project organizations structure the work program in the manner outlined here. Analmost-universal problem in these organizations is that a disproportionate share of both

96 William Ward

the routine and the nonroutine work is borne by the staff above the senior professionallevel. An often-quoted rule of thumb is that 80 percent of all work-routine andnonroutine-gets done by the top 20 percent of the staff. In a common anecdote, the topstaff do their own work during the day (predominantly nonroutine work) and do theirstaff's work at night (predominantly routine work). Most project organizations seriouslyneed to put in place systems and skills that will allow their organizations to better use alltheir human resources.

To improve the percentage of the operational load carried by the lowest-level staff(assistant level and professional level in table 6-2) requires that planners develop"established procedures" and that organizational development activities (includingtraining) closely correlate with the development of "standard operating procedures." It isimportant to understand that the development of training (in the present context,"project" training) for staff below the upper-middle-management level should go hand inhand with the development of SOPs. The standard model shows the SOPs beingdeveloped by upper-middle and middle management. The reality is that, in most cases,development of the SOPs and staff training should be a combined effort by trainers andsenior/middle management.

While tables 6-1 to 6-3 of course paint the world with broad and simple brushes, theynevertheless are useful in analyzing the worldwide system for meeting the human capitalneeds of project planning and implementation organizations in developing countries.

7

The Appraisal of Investment Projects: A TrainingApproach

Glenn P. Jenkins

The ultimate outcome of a public or a private investment hinges on the financial,economic, managerial, and political parameters associated with the project. A financial oreconomic net present value (NPV) calculated on a few sets of input variables tells onelittle about the potential of a project if its managerial capability is weak and its politicalsupport is fickle. While any analysis has its limitations, a project evaluation that closelylinks the financial and economic analysis, and in turn identifies the distribution ofbenefits and costs over the various interest groups, will be far more likely to identify fatalflaws in a project's design than will an appraisal that segments the analysis and does notaddress the interdependencies of the components.

A primary outcome of a project appraisal is the identification of aspects of the projectthat could inflict great damage on its performance. With this information the next step isto see if the project can be redesigned (organizationally, financially, or physically) tomake it more robust and resistant to external or internal shocks.

In order to carry out project evaluations in this way, it is our view that project analystsshould be comfortable with the skills of financial analysis, welfare economics, theevaluation of alternative sources of project uncertainty, and estimation of the distributionof costs and benefits across interest groups.

The Relationship between Financial and Economic Analyses

The economic analysis of an investment project is an essential complement to its financialappraisal. When traditional financial analysis examines the feasibility of the totalinvestment for a project, the appraisal is done from the point of view of the banker.Alternatively, if this project is a voluntary undertaking by its owners, analysts mustconsider the financial feasibility of the project from the owners' point of view.

These perspectives differ in the way they treat certain variables. For example, from anowner's point of view, a loan is a cash inflow and repayment of the loan and interest is acash outflow; from the banker's point of view, neither of these items is present in the cash

The assistance of Vivien Goldman and Mostafa Baher El-Hifnawi in the design andimplementation of the programs discussed here is greatly appreciated. Without their ideas anddedication, the outcomes of these "projects" would be very different.

97

98 Glenn P. Jenkins

flow. A close comparison of the annual (or even monthly) cash flows from these twoperspectives provides a means to analyze the way a particular set of financingarrangements distributes the risks of a project among its financial interests.

Traditionally, the financial internal rate of return (FIRR) and the NPV of the project arecalculated from the cash flow statements developed from these financial perspectives.These summary criteria do not, however, accurately assess the sustainability of a projector its riskiness. Consider a project that has both a large FIRR and a large positive NPV butalso has negative cash flows in the early years of its life. Such a project may go bankruptlong before it gets a chance to generate the large positive net cash flows expected in lateryears. Examination of the cash flow profile over the project's lifetime will indicate thesustainability and financial riskiness of the project.

An economic analysis of a project evaluates its feasibility from the point of view of thewhole country. A positive net present value indicates that the project will make a positivecontribution to the economic growth of the country as compared with utilizing the fundsfor their alternative uses. This criterion, however, is reliable only if the project isfinancially sustainable. If the financial structure of the project is such that it cannot pay itsfinancial costs, the potential net economic benefits have little probability of realization.

The economic values of both inputs and outputs differ from their financial valuesbecause of market distortions created by both the government and the private sector.Tariffs, export taxes and subsidies, excise and sales taxes, production subsidies, andquantitative restrictions are common distortions created by the government, whilemonopolies are a market phenomenon that can arise from either private or public sectoractions. Some market distortions are created by the public nature of the good or service.The values of common public services such as clean water, transportation, road services,and electricity are based on the maximum amount people are willing to pay for theservices. These values are often significantly greater than the financial prices people mustpay for the services. It is such factors that create divergences between the financial andthe economic prices for a project.

A consensus exists among accountants on the principles to use in undertaking afinancial appraisal of a potential investment, with relatively minor disagreements oncertain issues, such as the treatment of inflation. Also, financial analysts will usuallyagree on the cash flow and balance sheet requirements for a public or a private sectorproject to remain viable. Nevertheless, these accounting and financial principles are not asufficient guide for undertaking an economic appraisal of an investment.

The measurement of economic benefits and costs is built on the informationdeveloped in the financial appraisal, but in addition, it extensively uses the economicprinciples developed in the field of applied welfare economics (Harberger 1979). Thetechniques of economic investment appraisal are predicated upon three basic principlesof applied welfare economics: (a) the competitive demand price for a given unit of anitem measures the value of that unit to the demander (that is, his willingness to pay); (b)the competitive supply price for a given unit of a good or service measures the value ofthat unit to the supplier; and (c) when evaluating the net benefits or costs of a givenproject in a period, the costs and benefits accruing to each relevant group (for example, anation, tribe, profession, or an income class) should wherever possible be measured andidentified with the recipient, but in the assessment of a project's economic efficiency, allcosts and benefits should be added up to determine the overall net economic benefit ofthe project.

Appraisal of Investment Projects 99

At the present time, there is broad agreement in the field of applied investmentappraisal as to how to use these principles to estimate the economic prices of inputs andoutputs of projects. The alternative approaches do not differ significantly in theirtheoretical foundations and in the recommendations of approval or rejection (Ward,Deren, and D'Silva 1991, pp. 1-50). The choice among the different ways of expressingthe results of the analysis, however, depends on how analysts use the economic analysisin the overall financial, economic, and social distributional assessment of a project.

Traditional approaches to appraising investment projects have tended to undertakethe economic analysis in isolation from the financial analysis (Asian Development Bank1987; Dasgupta, Marglin, and Sen 1972; Little and Mirrlees 1979, 1982). But the frequentcollapse of development projects worldwide during the 1980s has forced project analyststo consider the sustainability of these investments. Furthermore, there continues to be aneed to identify the groups in society that will benefit from a project and those that willbear the costs. Finally, in an environment of trade liberalization and economic policyreform, the economic distortions that financially subsidize a project become some of themain sources of its financial risk.

Many project bankruptcies in the real estate development sector in Malaysia in the1980s arose from the government's cutback in subsidies to such projects. The reduction inthe rate of trade protection in Indonesia and Argentina provides two examples in whichthe realignment of financial prices of inputs and outputs with their fundamentaleconomic values caused a considerable decline in the financial profitability of someindustrial projects. Unless the prospective economic and financial performances arecompared on a period-by-period basis, it is difficult to assess a project's financial riskarising from the potential to either reduce or augment economic distortions.

The difference between the financial and the economic values of a good or servicerepresents a benefit or a cost that accrues to someone other than the financial sponsors ofthe project. Additional taxes cause financial values to decline, while some level ofgovernment will benefit from the tax revenues. If the financial price of a service, such aswater supply, is less than its economic value, this means that the consumers of waterreceive an increase in their real standard of living. A project that causes the price of agood or a service to fall, will create economic benefits that are greater than its financialrevenues. This difference between the financial and the economic values will represent adistributional gain to the consumers of the output, and a somewhat smaller loss to theother producers of the good or service who are competing in the market with the project.In each case, one can define the distributional effect by the difference between a financialprice and an economic value for the inputs or outputs of a project.

As international organizations start to stress cost recovery and the financialsustainability of projects, an important question for many public sector projects is thelevel of user charges that a project can levy on consumers while still maintainingdemand. Analysts can determine this only by estimating the economic value of thebenefits to the consumers as measured by their willingness to pay for the good or service.Again, a direct comparison of financial prices and economic values is essential beforesetting a financial charge that will enable the project to be financially sustainable.

To carry out an analysis of risk, distributional impact, and project sustainability, twoconditions must hold. First, we must express the financial and economic analysis in thesame units of account. This may be domestic currency at the domestic price level,domestic currency at the border price level, units of foreign exchange, or any other

100 Glenn P. Jenkins

monetary unit that is easy to measure and understand. When the financial analysis isdone in one unit of account and the economic analysis in another, the differencesbetween the financial and the economic values will have no meaning. Hence the elementsof risk and distribution become much more difficult to assess. Second, analysts shouldcompare the project's economic and financial profiles on a period-by-period basis andnot just summarize them in a single statistic, such as the net present value or the internalrate of return. Neither of these conditions has been stressed by the traditionalmethodologies of economic appraisal, but they are critical in the assessment of a project'slikelihood of survival, whether it be financially, economically, or politically.

The Choice of a Numeraire in the Economic Analysis of Projects

In theoretical developments of the economic analysis of projects, various authors haveused different numeraires to measure economic costs and benefits. Basically, they havemost frequently used three of them: (a) the willingness-to-pay numeraire expressed indomestic currency at the domestic price level, (b) the willingness-to-pay numeraireexpressed in domestic currency at the border price level, and (c) a numeraire by which allvalues of inputs and outputs are expressed in units of foreign exchange. Experts nowwidely recognize that the choice of a numeraire in conducting an economic analysis inisolation is largely a matter of convenience in measurement. (For an excellent discussionof the theoretical and practical features of each of these numeraires, see Ward, Deren, andD'Silva 1991, chapters 4-7.)

The financial analysis is generally done in domestic prices at the domestic price levelbecause these are both the "currency" and the "price level" in which the financialsponsors of the project operate. The use of any other numeraire quickly diminishes thelevel of understanding that the financial decisionmakers will derive from the results ofthe financial analysis.

In the past, various "schools" of project analysts have recommended using thenumeraires mentioned above. In particular, analysts following the method of the UnitedNations Industrial Development Organization or the approach developed by AmoldHarberger (1982) use domestic prices at the domestic price level as the numeraire, whilethose following the Little-Mirrlees approach use either domestic prices at the borderprice level or units of foreign exchange as the numeraire.

Using units of foreign exchange as the numeraire has a considerable theoreticalweakness when valuing the nontradable goods where the level of consumption(demand) is affected. In the shortcut approach to determine the economic value of aproject output in foreign exchange when the consumption of a nontradable is increased,analysts employ the technique of trying to find the foreign exchange value of thedecrease in consumption of a substitute tradable good. For example, they would measurethe value of additional electricity consumption by rural households by the foreignexchange saved through the reduction in their consumption of kerosene previously usedfor lighting. This approach leads to implausible results because, with the heat loss in thegeneration of electricity, the economy might use several times as much kerosene togenerate the electricity as the savings in kerosene consumption by the households; yetpeople are willing to pay many times more for electric lights than for kerosene lights.This type of problem will be typical of the valuation of substitute nontradable goods andservices.


Because of this defect, it would seem prudent to use as a numeraire either domesticprices at the domestic price level or domestic prices at the border price level. Theconversion of economic prices of inputs and outputs from one of these numeraires to theother is a trivial exercise. Clearly, if analysts want to undertake a modern risk andsustainability approach to project analysis, they should do both the financial and theeconomic analysis using the former numeraire. Employing either of them will have noeffect on the use of the internal rate of return or the net present value as summary criteriastatistics. At the same time, for those who wish to do a more complete analysis of risk,sustainability, and distributional assessment of a project, they need to express economicprices in domestic prices at the domestic price level so that the economic analysis can beeasily linked to analysis of the project's financial feasibility.

Organization of Courses in Investment Appraisal

For the past eight years, Harvard University has offered an eight-week course titled theProgram on Investment Appraisal and Management (PIAM). The course covers all theelements needed to complete a feasibility study of an investment project, including afinancial, economic, risk, and distributional analysis. It is taught through lectures andhaving students do eleven small case studies, make presentations of their work, andcomplete a comprehensive analysis of an actual project from their countries.

Participants use a spreadsheet program (until now LOTUS 1-2-3) to complete theirassignments, and they study microcomputers with the rest of the course. Each of thesmall case studies is designed to use progressively more sophisticated LOTUScommands. In addition, students conduct a Monte Carlo analysis of project variabilitywith a LOTUS-based program (RiskMaster) developed by Sawakis Sawides of theCyprus Development Bank.

The course is organized in three- or four-day cycles. Each module contains three orfour morning lectures, plus a small case that participants receive on the first or secondday. During the afternoons and evenings they spend approximately four hours a day ina computer laboratory to work on their case solutions.

Tutors, who continuously staff the computer laboratory, are completely familiar withthe case and the substance of the course in addition to having the necessary computerskills. It is critical that the computer laboratory, not the lecture hall, becomes the center oflearning. The lab is kept open, with at least two tutors, from 1 P.M. to 10 P.M. each day,including Sundays. On Saturdays it is open from 9 A.M. to 6 P.M.

The combination of microcomputer use and the learning process is complex andneeds careful consideration. We have become aware of some aspects of this process.Although these ideas may not be much more than hypothesis, it might be useful toconsider the following. First, for all participants the development of strongmicrocomputer skills employing a spreadsheet program is an important asset and amajor element for bringing about change in the way they will work and deal withmodernization in the future. Second, although there may be dozens of computers in aroom, the focus of participants on their own computer screen nevertheless gives them asense of seclusion so that they can concentrate and work efficiently in a pleasant, butcrowded, environment. Third, it is essential to deal with participants' frustrations withtheir lack of knowledge in the use of a spreadsheet program quickly. One minute is aslong as a participant should wait before a tutor or a colleague addresses his or her


problem. Fourth, portable computers should not be used in the first three or four weeksof a course. Participants who are shy about their rate of progress, or think they arecoming into the program with a lower level of computer skills than others, will have atendency to take the portable back to their hotel room to practice alone. Without thesupport of tutors and colleagues in the laboratory, these people will quickly fall farbehind and are potential early dropouts. Initially, they will also spend too much timeunproductively playing with the computer. Fifth, for demanding courses lasting morethan three weeks, it is not a good idea to have a computer available to every participantall the time. This will create a large number of early "burn outs." People need to berequired to take a break from the computer to read, think, and rest. If participants haveconstant access to a computer, they will tend to put in long hours in the first one or twoweeks and become tired or even ill. Fatigue has a tremendous negative impact on theireffectiveness in learning. Creating the need for definite "shifts" in the lab givesparticipants some sense of time organization, which they need. It also provides a solidjustification for the lab manager to require them to save their work and leave, even whenthey are on the brink of "solving" their intractable problem.

The day after participants receive the case, the class briefly discusses it to ensure thateveryone knows how to approach the problem and to introduce any new LOTUScommands that may be needed to solve the problem. About four days after receivingeach case, participants are required to submit their solutions in the morning, and twoparticipants are chosen to present their analyses. They are given about one hour's noticeto prepare their presentations. All assignments are graded and returned to the studentswith a model solution to the case.

During the first three weeks of the course, instructors cover the financial analysis ofprojects while at the same time presenting the foundations of economic analysis in aseries of one-hour morning lectures. Also during this time, the major cases are selectedand groups are assembled by the project managers. The first assignment, to prepare anoutline of their potential major case, is due the second day of the program. Interviews onthe major case topics start on day three. A tutor is assigned to each major case to monitorprogress and to assist if necessary.

The first three-week segment is completed with a quiz. Participants are informed thefirst day of the course that a quiz will be given in eighteen days. It is critical that thecourse commands the participants' attention and time from the first day of the program.People arrive from around the world with multiple agendas for their time; if they aregiven an opportunity to develop other plans or interests, it will be much harder, orimpossible, to build a consensus on the intellectual goals intended by the program.Breaks can be given throughout the course, but not at the beginning.

Following the first three-week segment, the next week's module deals with theevaluation of the causes of project variability and with methods of contracting to reducethe variability of project outcomes. Also during this week, the class has largely completedthe financial analysis of the major cases, and one of the shorter cases is completed.

The following three weeks cover the economic aspects of investment appraisal and thedistributive analysis of projects. In addition, five short cases and the major case are beingcompleted. Participants take a final quiz on Thursday of week seven and are then givenfour days (including a weekend) to finalize their major cases.

They submit all the major cases and make a series of presentations in the eighth weekof the program. Each case and its presentation are evaluated by the audience of their


peers using a one-page evaluation form. The participants are in turn evaluated on thequality of the comments they make on their case evaluations.

Finishing the program with the presentation of major cases (which are completed bygroups of three or four students) brings them back to the point of re-entry to their work.Many of the participants will be in a position to pass judgment on the projects that comefor their approval. The evaluations they make (eight to ten cases) will give participantssome experience in preparing their comments on proposals by using the knowledge theyhave acquired from the course.

This program is very intensive and labor intensive. The ratio of tutors to students isabout one to ten, and each tutor works in excess of forty hours a week. They make everyeffort to prevent students from dropping out. Individuals who seem to be having troubleand may quit get help without having to ask for it. Often such potential dropouts areassigned a particular tutor who becomes their close associate throughout the remainderof the course. It becomes that tutor's responsibility to make sure the person keeps upwith the work.

To reinforce their commitment to improving their work in this area, a very activealumni association exists. They distribute an alumni directory each year to previousprogram participants. This directory is widely used, and an extensive network has builtup around the world. A detailed outline of the program is presented in Appendix A.

Courses Given In-Country

The Harvard Institute for International Development has also been involved with shortercourses in project appraisal conducted in-country. These three-week courses are moresuccessful if combined with an advisory activity in which the participants evaluate aseries of actual projects after completing the course. It is preferable if these courses areresident in nature so that the participants are taken completely away from their work. Asyllabus of a typical course of this type appears in Appendix B.

In-country courses should begin by emphasizing the financial and risk analysiscomponents of investment appraisal, with a few lectures and cases to make theparticipants aware of the importance of economic variables. The way to get peopleenthused about investment appraisal is to make sure they are given strong skills. Withthese skills the participants will be viewed by their peers as people who will be able tomake changes in the way they do their work. The course participants will then have thecapability and motivation to experiment with the analysis, which makes the introductionof the economic concepts and variables much easier.

During the subsequent month to six weeks, when students are working on actualcases, they should hold formal meetings twice a week to consider any economicadjustments that have to be made to the financial cash flows of the project.

References

Asian Development Bank, Economics Office. 1987. Guidelinesfor the Economic Analysis of Projects.

Dasgupta, Partha, Stephen A. Marglin, and Amartya Sen. 1972. Guidelines for Project Evaluation.New York United Nations.


Harberger, Arnold C. 1979. "Three Basic postulates for Applied Welfare Economics: AnInterpretive Essay." Journal of Economic Literature 9 (September): 785-97

.1982. Project Evaluation: Collected Papers. Chicago, Illinois: University of Chicago Press.

Little, l. M. D., and J. A. Mirrlees. 1979, 1982. Project Appraisal and Planning for Developing Countries.New York: Basic Books.

Ward, William A., and Barry J. Deren with Emmanuel H. D'Silva. 1991. The Economics of ProjectAnalysis: A Practitioner's Guide. EDI Technical Materials. Washington, D.C.: World Bank.

Appendix ASyllabus for

Program on Investment Appraisal and Management (PIAM)June 22 - August 14,1992

Monday, June 22Morning Session

9:00 - 10:00 A.M. Course Opening Room 107, Pound HallAdministrationand Ms. Vivien GoldmanCourse Organization: Dr. Glenn JenkinsOverview ofCurriculum Dr. Roy KellyMajor Cases Mr. G. P. Shukla

10:00 - 10:30 A.M. Coffee Break

10:30 - 12:00 P.M. The Role of Project Dr. Glenn JenkinsEvaluation Room 107, Pound Hall

12:00 -1:00 P.M. Lunch

Aftemoon Session

1:00 - 2:30 P.M. Introduction to a Dr. Glenn JenkinsStrategy for the Room 101, Pound HallAppraisal ofInvestment Projects

Readings: PIAM Manual, chapters 1 and 2

2:30- 3:00 P.M. Coffee Break

3:00 - 5:00 P.M. Conceptual Dr. Glenn JenkinsFramework for the Room 107, Pound HallEvaluation ofIndustrial Projects

Readings: PIAM Readings, volume 1;Kuo, "A Conceptual Frameworkfor Economic Evaluation ofIndustrial Projects"; Roemer and Stem, The Appraisalof Development Projects, chapter 1

105

106

Tuesday, June 23

9:00 - 10:00 A. M. Economic Principles Dr. Roy KellyRoom 107, Pound Hall


10:30 - 12:30 P.M. Introduction to Mr. Baher El-HifnawiMicrocomputers Room 107, Pound Hall

12:30- 1:30 P.M. Lunch

1:30 - 3:30 P.M. Microcomputers: Mr. Baher El-HifnawiSpreadsheet Computer LaboratoryOperations 50 Church St., 4th FloorGroup A

3:45 - 5:45 P. M. Microcomputers: Mr. Baher El-HifnawiSpreadsheet Computer LaboratoryOperationsGroup B

Wednesday, June 24

9:00 - 10:00 A.M. Economic Principles Dr. Roy KellyRoom 107, Pound Hall

10:00 - 10:15 A M. Coffee Break

10:15 - 12:00 P.M. Development of Dr. Glenn JenkinsFinancial Cash Flows Room 107, Pound Hall

Readings: PIAM Manual,chapter 3, pp. 1-10;Henderson and Maness, chapters 1 and 3

Distribution of Case 1(Due Monday, June 29, 8:30 A.M.)

12:00 - 1:00 P.M. Lunch

1:00 - 3:00 P.M. Microcomputers Mr. Baher El-HifnawiGroup B Computer Laboratory

3:30 - 5:30 P.M. Microcomputers Mr. Baher El-HifnawiGroup A Computer Laboratory

107

Thursday, June 25



10:15 - 12:00 P.M. Operating Plan Dr. Glenn JenkinsRoom 107, Pound Hall

12:00 - 1:00 P.M. Lunch

1:00 - 3:00 P.M. Computerization of Mr. Baher El-HifnawiFinancial Cash Flows Computer LaboratoryGroup A

3:30 - 5:30 P.M. Computerization of Mr. Baher El-HifnawiFinancial Cash Flows Computer LaboratoryGroup B

Friday, June 26



10:15 - 12:00 P.M. Investment Plan Dr. Glenn Jenkins(continued) Room 107, Pound HallMeasurement ofCosts: Sunk orIncremental

12:00-2:00 P.M. Lunch

2:00 - 4:00 P.M. Progress with Profits: Mr. Richard PattenThe Development of Project Manager, HIIDRural Banking Bank Rakyat IndonesiaSystems Rural Banking Project

Monday, June 29



108

Distribution of Case 2(Due Friday, July 3, 8:30 A. M.)

10:15 - 11:00 A.M. Presentation of Case 1

Group A Dr. Roy KellyRoom 107, Pound Hall

Group B Mr. Baher El-HifnawiRoom 108, Pound Hall

11:00 - 12:00 P.M. Analysis of Financial Mr. G. P. ShuklaProfiles from Room 107, Pound HallAlternate Pointsof View

12:00 P. M. Lunch

Tuesday, June 30


10:00 -10:15 A.M. Coffee Break

10:15 - 12:00 P. M. Discounting and Mr. G. P. ShuklaAlternative Investment Room 107, Pound HallCriteriaNet Present Value

Readings: PIAM Manual,chapter 4, pp. 1-10;Roemer and Stem, volume 1,chapter 2, pp. 24-32

12:00 - 12:30 P. M. Discussion of Progresson Case 2Group A Tutors

Room 107, Pound HallGroup B Tutors

Room 108, Pound Hall12:30 P. M. Lunch

Wednesday, July 1


109

10:00- 10:15 A.M. Coffee Break

10:15 - 11:30 A.M. Adjusting Projects for Dr. Glenn JenkinsDifferent Lengths Room 107, Pound Hallof Life

11:30 - 12:30 P.M. Benefit-Cost Ratio, Dr. Glenn JenkinsPayback Period Room 107, Pound Hall

Readings: PIAM Manual, chapter 4,pp. 11-14. Roemer and Stem, volume 1,pp. 32-39

12:30 P. M. Lunch

Thursday, July 2



10:15 - 12:00 P.M. Internal Rate of Dr. Glenn JenkinsReturn Room 107, Pound Hall

Readings, PIAM Manual,chapter 4, pp. 14-21

12:00 p.m. Lunch

Friday, July 3


10:00 - 10:15 A. M. Coffee Break

10:15 - 12:45 P.M. Presentation of Case 2

Group A Mr. Baher El-HifnawiRoom 107, Pound Hall

Group B Mr. G. P. ShuklaRoom 108, Pound Hall

110

Distribution of Cases 3 and 4(Due Tuesday, July 7, 8:30 A M.)

12:45 - 2:00 P.M. Lunch

2:00 - 4:00 P.M. Lessons of the Mr. Marcelo SelowskiReform Process Chief Economistin Latin America Latin American and

Caribbean DeskWorld Bank

Sunday, July 5

12:00 - 1:00 P.M. Discussion of Progress TutorsCases 3 and 4 Computer Lab

3:00 - 4:00 P.M. Discussion Session TutorsRepeated Computer Lab

Monday, July 6

9:00 - 10:00 A.M. Foundations of Mr. Baher El-HifnawiRisk and Uncertainty Room 107, Pound Hall


10:30 - 12:00 P.M. Choice of Scale Dr. Shiv Kumarand Choice of Room 107, Pound HallProject Timing

Readings: PIAM Manual,chapter 5, pp. 1-8.

12:00 P. M. Lunch

Tuesday, July 7

9:00 - 10:00 A. M. Foundations of Risk Mr. Baher El-Hifnawiand Uncertainty Room 107, Pound Hall


10:30 - 11:30 A.M. Project Interdependencies Dr. Shiv KumarRoom 107, Pound Hall


111

Distribution of Cases 5 and 6(Due Friday, July 10, 8:30 A. M.)

11:30 - 12:30 P.M. Presentation of Cases 3 and 4

Group A Mr. Joe ThamRoom 107, Pound Hall


12:30-2:00 P.M. Lunch

2:00 - 4:30 P.M. A Natural Resource Professor David SmithInvestment Project: Vice DeanThe Negotiation Harvard Law SchoolProcess, Risk Room 107, Pound HallAllocations, andOptions for ContractualArrangements

Wednesday, July 8

9:00 - 10:00 A.M. Impact of Inflation on Mr. G. P. ShuklaInvestment Appraisal Room 107, Pound Hall


10:30 - 12:00 P.M. Impact of Inflation on Mr. G. P. ShuklaInvestment Appraisal Room 107, Pound Hall(continued)

Readings: PIAM Manual,chapter 6, pp. 1-7;Harberger, Project Evaluation,chapter 2, pp. 42-48

12:00 - 12:30 P. M. Discussion of ProgressCases 5 and 6

Group A TutorsRoom 107, Pound Hall

Group B TutorsRoom 108, Pound Hall

12:30-2:00 P.M. Lunch

112

2:00 - 4:00 P.M. The Economics of Theo PanayotouEnvironmental Institute FellowDegradation

Thursday, July 9

9:00 - 10:00 A. M. Inflation: Mr. G. P. ShuklaExamples of Impact Room 107, Pound Hall

Readings: PIAM Manual,chapter 6, pp. 7-15


10:15 -12:00 P.M. Discussion of Financial Mr. Baher El-HifnawiAnalysis of Inflation: Room 107, Pound HallA Case Study

12:00 P. M. Lunch

Friday, July 10

9:00 -12:00 P.M. Review Session Mr. Baher El-HifnawiRoom 107, Pound Hall

10:00 A.M. Group Picture

12:00-1:00 P.M. Presentation of Cases 5 and 6

Group A Mr. Joe ThamRoom 107, Pound Hall


1:00 P.M. Lunch

Saturday, July 11

9:00 - 10:00 A.M. PIAM Quiz 1 Room 107, Pound Hall

Monday, July 13

9:00 - 10:00 A.M. Introduction to Risk Dr. John EvansAnalysis and Room 107, Pound HallManagement

Readings: Brigham and others,"Risk Analysis and the Optimal Capital Budget"

113


10:30 - 12:30 P.M. Introduction to Monte Mr. Sawakis SawidesCarlo Analysis Cyprus Development

BankRoom 107, Pound Hall

Readings: Sawides,"Risk Analysis in ProjectAppraisal"

12:30 - 2:30 P. M. Lunch

Computer Application of Risk Room 107, Pound HallAnalysis Using RiskMaster

2:30 - 6:00 P.M. Group A Mr. Baher El-HifnawiComputer Lab

6:30 - 10:00 P.M. Group B Mr. Baher El-HifnawiComputer Lab

Tuesday, July 14

9:00 - 10:00 A.M. Illustration of Risk Dr. John EvansAnalysis and Room 107, Pound HallManagement: A CementAdditives Plant

Readings: Glenday,"Monte Carlo SimulationTechniques in the Valuation ofTruncated Distributions in theContext of Project Appraisal"


10:30 - 12:00 P.M. Topic continued Dr. John EvansRoom 107, Pound Hall

12:00-2:00 P.M. Lunch

2:00 - 6:00 P.M. Computer Application Mr. Baher El-Hifnawiof Risk Analysis Computer Lab

114

Distribution of Case 18Group A

Major CasesGroup B

Wednesday, July 15

9:00 - 10:00 A.M. Review of Quiz 1 Room 107, Pound Hall


10:30 - 12:00 P.M. Risk Bearing and Dr. John EvansChoice of Room 107, Pound HallContract Form

Readings: Blitzer,Lessard, and Paddock,"Risk Bearing and the Choiceof Contract Forms for OilExploration and Development"

12:00-2:00 P.M. Lunch

2:00 - 6:00 P.M. Computer Applications Mr. Baher El Hifnawiof Risk Analysis Computer Lab

Distribution of Case 19Group B

Major CasesGroup A

Thursday, July 16

9:00 - 10:00 A.M. Project Financing, Dr. John EvansFinancial Contracts, Room 107, Pound Halland Risks

Readings: Baldwin,Lessard, and Mason,"Budgetary Time Bombs: ControllingGovernment Loan Guarantees"


215

10:30 - 12:30 P.M. Topic continued Dr. John EvansRoom 107, Pound Hall

12:30-2:00 P.M. Lunch

2:00 - 6:00 P.M. Completion of Mr. Baher El-HifnawiRisk Cases

Group A Computer Lab

6:30 - 10:30 P.M. Completion of Mr. Baher El-HifnawiRisk Cases

Group B Computer Lab

Friday, July 17

9:00 - 10:00 A.M. Measurement of Dr. Glenn JenkinsCosts and Benefits in Room 107, Pound HallUndistorted Markets

Readings: PIAM Manual, chapters 7 and 8Optional: Harberger, "Three Basic Postulates ofApplied Welfare Economics: An Interpretive Essay,"volume 1

Distribution of Case 9(Due Wednesday, July 22, 8:30 A.M.)

10:00 - 10:15 AM. Coffee Break

10:15 - 11:30 P.M. Topic continued Dr. Glenn JenkinsRoom 107, Pound Hall

11:30 - 12:30 P.M. Presentation of RiskAnalysis Cases

Group A Dr. John EvansRoom 107, Pound Hall

Group B Mr. Sawakis SawidesRoom 108, Pound Hall

12:30 - 2:00 P.M. Lunch

2:00 - 4:00 P.M. Marketing Analysis Mr. Sawakis Sawidesin Project Evaluation Room 107, Pound Hall

116

Readings: Sawides,"Marketing Analysis in ProjectEvaluation"

Monday, July 20

9:00 - 10:00 A.M. Measurement of Dr. Roy KellyBenefits and Costs Room 107, Pound Hallin Distorted Markets(Taxes and Subsidies)

Readings: PL4M Manual,chapter 9, pp. 1-14


10:15 - 12:00 P.M. Measurement of Dr. Roy KellyBenefits and Costs Room 107, Pound Hallin Distorted Markets(Taxes and Subsidies)

12:00 P. M. Lunch

Tuesday, July 21

9:00 -10:00 A. M. Ceiling Prices and Dr. Roy KellyRationing Room 107, Pound Hall

Distribution and Discussion ofCase 10, Part 1, in Class



10:15 - 12:30 P.M. Monopoly and Dr. Roy KellyMeasurement of Room 107, Pound HallCosts and Benefits

Discussion of Case 10,Part 2, in Class


117

12:30 - 2:00 P.M. Lunch

2:00 - 4:00 P.M. Review Session Room 107, Pound Hall(Nontradable Goods)

Wednesday, July 22

9:00 - 10:00 A.M. Economic Prices for Dr. Roy KellyTraded Goods Room 107, Pound Hall


Distribution of Case 11(Due Monday, July 27, 8:30 A. M.)




Group B Mr. Joe ThamRoom 108, Pound Hall

12:00-2:00 P.M. Lunch

2:00 - 4:00 P.M. Performance Contracts: Dr. Praja TrevediAn Approach to Professor of PublicImproving EnterprisePublic Enterprise Indian Institute ofPerformance Management

Thursday, July 23

9:00 - 10:00 A.M. Economic Prices for Dr. Roy KellyTraded Goods Room 107, Pound Hall(continued)


10:15 - 12:00 P.M. Topic continued

12:00 - 12:30 P.M. Discussion of Progress Case 11

118

Group A TutorsRoom 107, Pound Hall

Group B TutorsRoom 108, Pound Hall

12:30 P. M. Lunch

Friday, July 24

9:00 - 10:00 A. MN. The Economic Price of Dr. Roy KellyForeign Exchange Room 107, Pound Hall


10:15 - 12:00 P. . The Economic Cost of Dr. Roy KellyTraded Goods and Room 107, Pound Hallthe Shadow Priceof Foreign Exchange

Distribution of Case 12(Due Wednesday, July 29, 8:30 A. M.)


12:00 P. M. Lunch

Monday, July 27

9:00 - 10:00 A.M. Estimation of Dr. Roy KellyEconomic Prices Room 107, Pound Hallwith More ThanOne Distorted Market

Readinigs: PIAM Manual,chapter 11, pp. 14.


10:15 - 11:30 A. M. Topic continued




119

12:30 - 1:30 P.M. Lunch

1:30 - 3:30 P.M. Project Finance Mr. Adebayo OgunlesiVice President ofProject FinanceFirst Boston Corporation

Tuesday, July 28

9:00 - 10:00 A.M. Estimation of Benefits Dr. Roy Kellyof Impact in Room 107, Pound HallComplementary andSubstitute Goods



10:15 - 12:00 P.M. The Economic Dr. Glenn JenkinsOpportunity Cost Room 107, Pound Hallof Public Funds

Readings: PIAM Manual, chapter 12;Harberger, "On Country Risk andthe Social Cost of Foreign Borrowing byDeveloping Countries," volume 1

12:00 P.M. Lunch

Wednesday, July 29

9:00 - 10:00 A.M. The Economic Dr. Glenn Jenkins,Opportunity Cost Room 107, Pound Hallof Public Funds(continued)

Readings: Little and Mirrlees,"Project Appraisal and PlanningTwenty Years On," volume 1

Distribution of Case 13(Due Monday, August 3, 8:30 A.M.)


120

10:15 - 11:30 A. M. Topic continued


Group A Mr. G. P. ShuklaRoom 107, Pound Hall


12:30 - 2:00 P.M. Lunch

Thursday, July 30

9:00 - 10:00 A.M. Distributive Analysis Dr. Glenn JenkinsRoom 107, Pound Hall


10:15 - 12:00 P.M. Distributive Analysis(continued)

Distribution and Discussion of Case 16

12:00 - 12:30 P.M. Discussion of Progress Case 13

Group A TutorsRoom 107

Group B TutorsRoom 108

12:30 - 2:00 P.M. Lunch

2:00 - 3:30 P.M. Review Session(Traded Goods and Foreign Exchange)

Friday, July 31

9:00 - 10:00 A.M. Where is Professional Arnold HarbergerProject Appraisal? Department of

Economics,University of ChicagoandUniversity ofCalifornia at Los Angeles

121

Readings: PIAM Readings, volume 1;Harberger, "Reflections on Social Project Evaluation"


10:15 - 12:00 P.M. Topic continued

12:00 - 2:00 P.M. Lunch

4:00 - 6:00 P.M. Reception Faculty Club20 Quincy Street

Monday, August 3

9:00 - 10:00 A. M. Social-Distributive Dr. Glenn JenkinsAppraisal of Projects Room 107, Pound Hall

Readings: Harberger, "BasicNeeds and Distributional Weights inSocial Cost Benefit Analysis" and "Onthe Use of Distributional Weights inSocial Cost Benefit Analysis," volume 1


10:15 - 11:30 A.M. Topic continued


Group A Mr. Baher El-HifnawiRoom 107, Pound Hall

Group B Mr. Joe ThamRoom 108, Pound Hall

12:30 - 2:00 P.M. Lunch

2:00 - 4:30 P.M. Review Session Room 107, Pound Hall(optional) on Risk,Secondary Effects,and Cost of Capital

Tuesday, August 4

9:00 - 10:00 A.M. The Economic Dr. Glenn JenkinsOpportunity Cost Room 107, Pound Hallof Rural Labor

122

Readings: PIAM Manual, chapter 13, pp. 1-6;Harberger, "Project Evaluation," chapter 7

Distribution of Case 14


10:15 - 12:00 P.M. The Economic Dr. Glenn JenkinsOpportunity Cost Room 107, Pound Hallof Rural Labor(continued), includingsolution of Case 14

12:00 P. M. Lunch

Wednesday, August 5

9:00 - 10:00 A. M. The Economic Dr. Glenn JenkinsOpportunity Cost Room 107, Pound Hallof Urban Labor

Readings: PIAMAManual,chapter 13, pp. 7-10


10:15 - 12:00 P.M. The Social Dr. Glenn JenkinsOpportunity Cost of Room 107, Pound HallUrban Labor

Distribution and Discussion of Case 15

Readings: PIAM Readings,volume 1; Harberger,"The Social OpportunityCost of Labor: Problems ofConcept and Measurement As Seenfrom a Canadian Perspective"

12:00 - 2:00 P.M. Lunch

2:00 - 4:00 P.M. Quiz Review Session Rooms 107 and 108,Pound Hall

123

Thursday, August 6

2.00 - 4:00 P.M. Quiz 2 Room 107, Pound Hall

Friday, August 7

Preparation of Major Cases

Monday, August 10

Preparation of Major Cases

All Major Cases Due in Computer Lab at 10:00 P.M.

Tuesday, August 11Morning Session

9:00 - 10:15 A.M. Major Case Presentations:

Case 1 Room 107, Pound Hall



10:45 - 12:00 P.M. Major Case Presentations:



12:00-1:30 P.M. Lunch

Afternoon Session




2:45 - 3:00 P.M. Coffee Break



124


4:15-4:30 P.M. Coffee Break

4:30 - 5:45 P.M. Case 9 Room 107, Pound Hall


Wednesday, August 12Morning Session

9:00 - 10:15 A. M. Major Case Presentations:







12:00 P. M. Lunch

Afternoon Session









125




Thursday, August 13

6:00 P. M. Certificate Presentation Faculty ClubReception and Dinner

126

Appendix BSyllabus for

Project Appraisal and Risk Analysis ManagementJanuary 6 - 25, 1992

Week 1

Monday, January 6

09:00 - 12:00 P.M. Course Opening

12:00 - 13:45 P.M. Lunch

13:45 - 15:00 P.M. Components of Project Evaluation

Readings: PIAM: Jenkins and Harberger,Cost-Benefit Analysis of InvestmentDecisions, chapters 1 and 2, pp. 2.1 - 2.5

15:00 - 15:30 P. M. Coffee Break

15:30 - 17:00 P.M. The Components of Cash Flow Analysis

Readings: Manual,chapter 2, pp. 2.5 - 2.16chapter 3, pp. 3.1 - 3.6

Tuesday, January 7

08:30 - 10:00 A.M. Analysis of Cash Flows

Readings: Manual, chapter 3, pp 3.1 - 3.10;Henderson and Maness, The FinancialAnalyst's Deskbook, chapters 1 and 3


10:30 - 12:00 P.M. Continuation of Topic

12:00 - 13:45 P.M. Lunch

13:45 P.M. - onward Distribution and Analysis of Case I(Due Thursday, January 9, 8:30 A. M.in Computer Laboratory)Analysis of Case I

127

Wednesday, January 8

08:30 - 10:00 A. M. Evaluation of Investment Projectsfrom Alternative Points of View

Readings: Matnual,

chapter 3, pp. 3.10 - 3.20


10:30 - 12:00 P.M. Discounting and Other Investment Criteria

Readings: Manual: chapter 4, pp. 4.1 - 4.12

12:00 - 13:45 P.M. Lunch

13:45 P.M. - onward Analysis of Case IFinancial Cash Flows(Computer Laboratory)

Thursday, January 9

08:30 - 10:00 A. M. Continuation of Discounting andOther Investment Criteria

Readings: Manual, chapter 4, pp. 4.13 - 4.19


10:30 - 12:00 P.M. Presentation of Case I by Participants

12:00 - 13:45 P.M. Lunch

13:45 P.M. - onward Distribution and Analysis of Cases II A and II B(Due Monday, January 13, 8:30 A. M.,in Computer Laboratory)

Friday, January 10

08:30 - 10:00 A.M. Determination of Optimal Scaleof Projects

Readings: Manual, chapter 5, pp. 5.1 - 5.4


128

10:30 - 12:00 P.M. Determination of Optimal Timing ofProjects and the Analysis of SeparableProject Components

Readings: Manual, pp. 5.4 - 5.13

Distribution of Case III(Due Tuesday, January 14, 8:30 A. M.)

12:00- 14:30 P.M. Lunch

14:30 P.M. - onward Analysis of Case II(Computer Laboratory)

Saturday, January 11

08:30 - 10:00 A.M. Analysis of Rehabilitation orTermination Situation

Readings:The Case of the Limassol JuiceCompany


10:30 - 11: 30 A. M. Continuation of Topic

11:30 - 12:00 P.M. Discussion of Scale Aspect of Case III

12:00 - 13:45 P. M. Lunch

13:45 P.M.- onward Analysis of Cases II & III(Computer Laboratory)

Week 2

Monday, January 13

08:30 - 10:00 A.M. Discussion of Time Aspect of Case III andPresentations of Cases II A and II B

10:00- 10:30 A.M. Coffee Break

10:30 - 12:00 P.M. Completion of Case III(Computer Laboratory)

12:00 - 13:45 P.M. Lunch

129

13:45 P.M.- onward Completion of Case III(Computer Laboratory)

Tuesday, January 14

08:30 - 10:00 A.M. Impact of Inflation on Investment Projects

Readings: Manual, chapter 6


10:30 - 11:30 A.M. Impact of Inflation on Investment Projects (continued)

11:30 - 12:00 P.M. Presentation of Case III (5A and 5B)

12:00 - 13:45 P.M. Lunch

13:45 - 15:30 P.M. A Case Study of Inflation Impacts onProjects

15:30 P.M. - onward Distribution and Analysis of ComprehensiveCases (Computer Laboratory)


08:30 - 10:00 A. M. Statistical Foundations of Risk Analysis


10:30 - 12:00 P.M. Completion of Comprehensive Cases

12:00 - 13:45 P.M. Lunch

13:45 P.M. - onward Completion of Comprehensive Cases

Thursday, January 16

08:30 - 10:00 A.M. Impact of Economic Factors on ProjectFeasibility: An Overview

Readings: Kuo, "A Conceptual Frameworkfor the Economic Evaluation ofIndustrial Projects"


130


12:00 - 13:45 P.M. Lunch

13:45 P.M. - onward Completion of Comprehensive Cases

Friday, January 17

08:30 - 10:00 A.M. Completion of Comprehensive Cases



12:00- 14:30 P.M. Lunch

14:30 - 16:00 P.M. Presentation of Comprehensive Case I

16:00 - 17:30 P.M. Presentation of Comprehensive Case II

Saturday, January 18

08:30 - 10:00 A.M. Presentation of Comprehensive Case III


10:30 - 11:45 A.M. Presentation of Comprehensive Case IV

11:45 - 13:00 P.M. Presentation of Comprehensive Case V

13:00-14:30 P.M. Lunch

FREE

Week 3

Monday, January 20

08:30 - 10:00 A. M. Analytical Foundations of RiskAnalysis


10:30 - 12:00 P.M. Risk and Project Analysis:Introduction to Sensitivityand Risk Analysis

131

Readings: Sawakis Sawides, "Risk Analysis inInvestment Appraisal"Optional: Pouliquen, "Risk Analysis in ProjectAppraisal"

12:00 - 13:45 P.M. Lunch

13:45 P.M. - onward Introduction to RiskMaster

Tuesday, January 21

08:30 - 10:00 A. M. Principles of Contracting, Risk Sharing, andRisk Reduction

Readings: Blitzer, C., D. Lessard, and J. Paddock, 1984."Risk-Bearing and the Choice ofContract Forms for Oil Explorationand Development," The Energy Journal5(1): 1-27.

Glenday, G. 1989. "Monte CarloSimulation Techniques in theValuation of TruncatedDistributions in the Context ofProject Appraisal." HarvardInstitute for InternationalDevelopment, June 12.

Optional:Lessard, D. R. 1987. "The FinancialComponent of Foreign DirectInvestment." In Costs and Benefits, pp. 131-52.Richard D. Robinson, ed., New York:Praeger Publishers.


10:30 - 12:00 P.M. Continuation of Contracting, Risk Sharing,and Risk Reduction

12:00 - 13:30 P. M. Lunch

13:30 P.M. - onward Application of Risk Analysis toComprehensive Cases

132


08:30- 10:00 A.M. Project Financing

Readings: Beale, C. W. 1984. "Trends in Non-RecourseFinancing." In First Boston Corporation.New York.

Niehuss, J. M. 1982. "An Introduction toInternational Project Financing." In Robert Hellawell andDon Wallace, Jr., eds., NegotiatingForeign Investnments: A Manualfor the ThirdWorld, Washington, D. C.: International Law Institute.


10:30 - 12:00 P.M. Completion of Application of Risk Analysisto Comprehensive Cases

12:00 - 13:30 P. M. Lunch

13:30 P.M. - onward Completion of Application of Risk Analysisto Comprehensive Cases

Thursday, January 23

08:30 - 10:00 A.M. Quiz


10:30 A.M. - onward Completion of Application of Risk Analysisto Comprehensive Cases

Friday, January 24

08:30 - 09:30 A. M. Presentation of Cases: Session I09:30 - 10:30 A.M. Session II

10:30- 10:45 A.M. Coffee Break10:45 - 12:00 P.M. Session III12:00 - 14:30 P. M. Lunch14:30 - 16:30 P.M. Session IV

16:30 - 18:00 P.M. Session V

Distributors of CZECH REPUBLIC R O.Y International Oxford University Press Mundi-Prensa Barcelona

World Bank National lntormation Center PC Box 13056 5 Bangalore Town Consell de Cent. 391

Publications prodelna, Konviktska 5 Tel Aviv 61130 Sharae Faisal 08009 BarcelonaPrices and credit terms CS - 113 57 Prague 1 PO Box 13033narynfrom country to DENMARK Palestinian Authonty/Middle Karachi-75350 SRI LANKA

counryf. Consult your SamfundsLitteratur East THE MALDIVESlocal distributor before Rosenoerns Ala 11 Index Information Services PERU Lake House Bookshopplacing an order. DK-1970 Fredenksberg C PO.B. 19502 Jerusalem Ediloral Desarrollo SA PO. Box 244

Apartado 3824 1 0O, Sir Chittampalam A.

ALBANIA EGYPT, ARAB REPUBLIC ITALY Lima 1 Gardiner Mawatha

Adrion Ltd. OF Lcosa Commissionana Colombo 2Pedat Rexhepi Str Al Ahram, Al Galaa Street. Sansoni SPA PHILIPPINESPai. 9, Shk. 1, Ap. 4 Cairo Via Duca Di Calabna, 1/1 Intemational Booksource SWEDENTirana Casella Postale 552 Center .Inc. Fntzes Customer Serice

The Middle East Observer 50125 Firenze Suite 720, Cityland 10 Regeingsgaton 12

ARGENflNA 41, Shent Street, Cairo Condominium Tower 2 S-106 47Oictna del LUbro Intemacional JAMAICA H.V dela Costa, comer StockholmAv Cordoba 1877 FINLAND Ian Randle Publshers Ltd. Valero St., Makati, Metro

20BuerdosAires Akeleeminen Kirlakauppa 206 Old Hope Road. Manila Wennergren-Williams AB1120 Buenos Aires ~P0. Bos 23 Kingston 6 P.0. Box 1305

AUSTRAUA, FIJI, PAPUA FIN-00371 Helsinki POLAND S-171 25

NEW GUINEA, SOLOMON JAPAN Intemraional Publishing Solna

ISLANDS, VANUATU, AND FRANCE Eastem Book Service ServiceWESTERN SAMOA Wold Bank Publications Hongo 3-Chome. Ul. Piekna 31/37 SWITZERLAND

D.A. lntomiation Services 66, avenue d'Idna Bunkyo-ku t13, Tokyo 00-57 Warzawa Ubraine Payot

648 Whitehorse Road 75116 Pans Service InstilubonnelMitcham 3132 KENYA PORTUGAL C6tes-de-Montbenon 30Vctona GERMANY Atnca Book Service (E.A.) Livrana Portugal 1002 Lausanne

UNO-Venag Ltd. Rua Do Carmo 70-74

AUSTRIA Poppelsdorter Allee 55 Ouaran House, Mtangano 1200 Usbon Van Diermen Editons

Gerold and Co. 53115 Bonn Street Tachniues

Graben 31 P0O. Box 45245. Nairobi ROMANIA Ch. de Lacuez 41A-lOll Wien GREECE Compani De Librari CH18B07 Blossy

Papasotbou S.A. KOREA, REPUBUC OF Bucuresb S.A.

BANGLADESH 35, Stoumara Str, 106 82 Daepn Trading Co. Ltd. Str. Upscana no. 26, sector 3 TANZANIAMicro Industries Development Athens PO. Box 34, Yeosida Bucharesl Oxford University PressAssistance Society IMIDAS) Seoul Maktaba StreetHouse S. Road 16 HONG KONG, MACAO RUSSIAN FEDERATION PO Box 5299Dhanmondi RlArea Asia 2000 Ltd. MALAYSIA Isdatelstvo Ves Mir> Dar es SalaamDhaka 1209 Sales & Circulation University at Malaya 9a, Kolpachniy Pereslok

Department CooperativeBbkshop, Moscow 101831 THAILANDBELGIUM Seebird House, unit 1101-02 Limded. PO. Box 1127 Central Books DistriburortJean De Lannoy 22-28 Wyndham Street Jalan Pante Baru SAUDI ARABIA, QATAR 306 Shorn RoadAv. du Roi 202, 1060 Central, Hong Kong 59700 Kuala Lumpur Jair Book Store BangkokBnussels HUNGARY MEXICO P.O. Box 3196, Riyadh 11471

BRAZIL Foundation for Market INFOTEC SINGAPORE, TAIWAN, JAMAICAPubicacdes Tecntcas Economy Apattado Postal 22860 MYANMAR, BRUNEI Systemati Stdes Unit

Interacioais tda.Dombovan Ut' 17-1 9 i0 Tap,Me DF Asahgate Publishing Asia 09 Welts StreetRua Peixoto Gornide, 209 H-1117 Budapest Pacific Pte. Ltd. Curepe01409 Sao Paulo, SP. INDIA NETHERLANDS 41 KaJlang Pudding Road Trinidad

CANADA ~~~Allied Publishers Ltd. De LUndeboonVinOr- #04-3 West ladies

Renout Publishing Co Ltd 7. Mounta R u 0 0Boa2s Gotden Wheel Building1294 Algoma Road 7460~~~~~~~~P.O Aox H202 ege Singapore 349318 UGANDA

Ott4Awgoa RodMda-6000740AHakern Gustro Ltd.Otaawa INDONESIA SLOVAK REPUBUC Madhvani BuikngOntano K1B 3W8 Pt. Indira Lmited NEW ZEALAND Slovart G.T.G. Ltd. PO Box 9997

CHINA Jalan Borobudur 20 tmvaEBSCO NZ Ltd. Krupinska 4 Plot 16/4 Jinja Rd.CHINA JalanP.O Boxobu181 2 Privale Madl Bag 99914 P0 Boa 152 Karmpala

Chna Financial & Economic RO. Box 181 New Ma0ket 85299 Bra1islava Publishing House Jakana 10320 Auckland UNITED KINGDOM8, s oi n ng J IRAN SOUTH AFRICA, Microinto Ltd.Beijing Kowkab Publishers NIGERIA BOTSWANA PO. Box 3

COLOMBIA PO. Box 19575-511, Tehran Untvrsity Press Lmited For single tides: Hamton,Iritoenlace LIda. ~~~~~~~~~~Three Crowns Buiting Oxford University Press Hampshire GU34 2PG

Apartadio Aereo 34270 Ketab Bars Co. Publishers JecoSouthrem Atrica EnglanBogotc DE. Khaled Eslamboli Ave1th St Private Mad Bag 5 PO. Box 1141

Kusheh Delatrooz No. 8 Ibadan Cape Town 8000 ZAMBIA

COTE D'IVOIRE Tehran NORWAY University bookahopCentre d'Edibon el d ~~NRWAYFor subscnption onlcers.: Great East Road Campus

Centre dEditio Naetn demaio eneDitfusion Atnicalnes (CEDA) IRELAND Naivesen Intonnation Center Internationail Subscription P.O. Box 3237904 BRP 541 Government Supplies Agency Book Department Sv.Lusakaa

AbW04 Plateau Oifig an tSolitlhair R.O. Box 6125 Etterstad P.O. Boa 410954-5 Harcourt Road N4; Oslo 6 Craighall ZIMBABWE

CYPRUS Dublin 2 PAKISTAN Johannesburg 2024 Longman ZimbabweCenter of Applied Ressrchl IRE iz okAec PI (Pta.) Ltd.Cyprus CollegeRE iraBo gec PI Tourte Road, Arterisl6, Diogenes Street, Engomi Yozmot Literature Ltd. 65, ShahraheOuaid+Azam Mundi-Prensa Libros, S.A. PO. Boa ST125

PO. Boa 2006 P0. Boa 55055 P.O. Boa N Castello 37 SouthertDnNicsia Tet Aviv 81580 Lahtore 540D0 28001 Madrid

1 The World Bank

I !1|1~~

* ,-0Z

* _ * I

* . I , 0 i

* . 0 ; 1

* . BS 1

SG

OS 0i

jfII!jrj II IIIa&&Iiii II 535 023E924 9

ISBN 0-8213-3325-9

Documents

The Economic Evaluation of Projects Economic App.pdf(Project Evaluation for the Next Decade and Notes on Some Issues in Social Project Evaluation), Glenn Jenkins (The Appraisal of